C259etukansi.fm Page 1 Tuesday, November 28, 2006 9:55 AM
C 259
OULU 2006
UNIVERSITY OF OULU P.O. Box 7500 FI-90014 UNIVERSITY OF OULU FINLAND
U N I V E R S I TAT I S
S E R I E S
SCIENTIAE RERUM NATURALIUM
Professor Mikko Siponen
HUMANIORA
Professor Harri Mantila
TECHNICA
Professor Juha Kostamovaara
ACTA
U N I V E R S I T AT I S O U L U E N S I S
Hannu Sorvoja
E D I T O R S
Hannu Sorvoja
A
B
C
D
E
F
G
O U L U E N S I S
ACTA
A C TA
C 259
NONINVASIVE BLOOD
PRESSURE PULSE DETECTION
AND BLOOD PRESSURE
DETERMINATION
MEDICA
Professor Olli Vuolteenaho
SCIENTIAE RERUM SOCIALIUM
Senior Assistant Timo Latomaa
SCRIPTA ACADEMICA
Communications Officer Elna Stjerna
OECONOMICA
Senior Lecturer Seppo Eriksson
EDITOR IN CHIEF
Professor Olli Vuolteenaho
EDITORIAL SECRETARY
Publications Editor Kirsti Nurkkala
ISBN 951-42-8271-X (Paperback)
ISBN 951-42-8272-8 (PDF)
ISSN 0355-3213 (Print)
ISSN 1796-2226 (Online)
FACULTY OF TECHNOLOGY,
DEPARTMENT OF ELECTRICAL AND INFORMATION ENGINEERING,
INFOTECH OULU,
UNIVERSITY OF OULU
C
TECHNICA
ACTA UNIVERSITATIS OULUENSIS
C Te c h n i c a 2 5 9
HANNU SORVOJA
NONINVASIVE BLOOD PRESSURE
PULSE DETECTION AND BLOOD
PRESSURE DETERMINATION
Academic dissertation to be presented, with the assent of
the Faculty of Technology of the University of Oulu, for
public defence in Raahensali (Auditorium L10), Linnanmaa,
on December 8th, 2006, at 12 noon
O U L U N Y L I O P I S TO, O U L U 2 0 0 6
Copyright © 2006
Acta Univ. Oul. C 259, 2006
Supervised by
Professor Risto Myllylä
Reviewed by
Professor Pasi Karjalainen
Professor Tadeusz Pustelny
ISBN 951-42-8271-X (Paperback)
ISBN 951-42-8272-8 (PDF)
http://herkules.oulu.fi/isbn9514282728/
ISSN 0355-3213 (Printed)
ISSN 1796-2226 (Online)
http://herkules.oulu.fi/issn03553213/
Cover design
Raimo Ahonen
OULU UNIVERSITY PRESS
OULU 2006
Sorvoja, Hannu, Noninvasive blood pressure pulse detection and blood pressure
determination
Faculty of Technology, University of Oulu, P.O.Box 4000, FI-90014 University of Oulu, Finland,
Department of Electrical and Information Engineering, Infotech Oulu, University of Oulu, P.O.Box
4500, FI-90014 University of Oulu, Finland
Acta Univ. Oul. C 259, 2006
Oulu, Finland
Abstract
This thesis describes the development of pressure sensor arrays and a range of methods suitable for
the long-term measurement of heart rate and blood pressure determination using a cuff and a pressure
sensor array on the radial artery. This study also reviews the historical background of noninvasive
blood pressure measurement methods, summarizes the accuracies achieved and explains the
requirements for common national and international standards of accuracy.
Two prototype series of pressure transducer arrays based on electro-mechanical film (EMFi) were
designed and tested. By offering high (∼TΩ) resistance, EMFi is an excellent material for low-current
long-term measurement applications. About 50 transducer arrays were built using different
configurations and electrode materials to sense low-frequency pressure pulsations on the radial artery
in the wrist. In addition to uniform quality, essential requirements included an adequate linear
response in the desired temperature range. Transducer sensitivity was tested as a function of
temperature in the range of 25–45 °C at varying static and alternating pressures. The average
sensitivity of the EMFi used in the transducers proved adequate (∼2.2 mV/mmHg and ∼7 mV/mmHg
for normal and high sensitive films) for the intended purpose.
The thesis also evaluates blood pressure measurements by the electronic palpation method (EP)
and compares the achieved accuracy to that of the oscillometric method (OSC) using average intraarterial (IA) blood pressure as a reference. All of these three measurements were made
simultaneously for each person. In one test group, measurements were conducted on healthy
volunteers in sitting and supine position during increasing and decreasing cuff pressure. Another
group, comprising elderly cardiac patients, was measured only in the supine position during cuff
inflation. The results showed that the EP method was approximately as accurate as the OSC method
with the healthy subjects and slightly more accurate with the cardiac patient group. The advantage of
the EP method is that also the wave shape and velocity of arterial pressure pulses is available for
further analysis, including the assessment of arterial stiffness.
Keywords: blood pressure monitoring, cuff, pressure transducer, pulse transit time, pulse
wave velocity, radial artery pulsation
Acknowledgements
Focusing on cross-scientific research on blood pressure measurement methods, the work
leading to this thesis was carried out mainly during 1993-2002 in the Department of
Electrical and Information Engineering at the University of Oulu.
First of all, I would like to thank Prof. Risto Myllylä for encouraging me in various
ways to finalize the thesis and also for arranging time to do it.
I also wish to express my deep gratitude to my parents, Kaarlo (deceased) and Linnea
Sorvoja, and my sisters and brothers, who have offered unstinting support during my
whole life. My father’s early death in 1992 due to a cardiovascular condition provided the
motivation to focus on biomedical engineering research, when it was time to start my
Diploma Thesis.
I owe my warmest thanks to my wife Irene, her children and grandchildren. They have
brought meaning and joy into my life.
I am grateful to Polar Electro Ltd. for funding the projects and to the Tauno Tönning
Foundation for supporting this work through two grants. I also wish to express my thanks
to personnel of the Oulu University Hospital and Northern Ostrobothnia Hospital District,
especially Päivi Kärjä-Koskenkari, Juha Koskenkari, Mauno Lilja and Y. Antero
Kesäniemi, for providing me the opportunity to conduct the necessary measurements at
the hospital.
Moreover, I wish to thank Seppo Nissilä, Vuokko-Marjut Kokko, Jukka Hast, Antti
Ruha, Eija Vieri-Gashi, Jari Miettinen, and Mika Sorvisto, previous co-workers, for
participation in the research.
I would also like to convey my appreciation to the entire personnel in the
Optoelectronics and Measurement Techniques Laboratory for creating a pleasant working
environment. My fellow-workers in the Electronics Laboratory and the whole staff at the
department’s workshop deserve the warmest thanks for making the mechanical
components for the measurement system.
Last, but not least, I also wish to extend my sincerest thanks to my friends, who have
remained close to me since my student days.
Oulu, October 2006
Hannu Sorvoja
List of terms, symbols and abbreviations
Ω
£
°C
A
C
Hz
m
mmHg
s
V
W
Ohm
Pound
degrees Celsius
Ampere
Coulomb
Hertz
meter
millimeter of mercury, unit for pressure, 133 Pascals
second
Volt
Watt
α
ρ
τ
µ
c
CR
constant depending on arterial properties
symbol for blood density
symbol for time constant
micro, 10-6
symbol for the velocity of blood pressure pulses
symbol for characteristic ratio of the diastolic or systolic determination ratio of
the maximum pulse amplitude in the oscillometric blood pressure measurement
method
symbol for thickness of an artery
symbol for derivative of pressure
mathematical constant (Euler), base of the natural logarithm, 2.71828…
symbol for modulus of elasticity
mill, 10-3
pico, 10-12
symbol for pressure
symbol for correlation coefficient
tera, 1012, temperature, (also Tesla)
symbol for output voltage
d
dp/dt
e
E
m
p
p, P
R
T
Uout
A/A
accuracy classification grade in the AAMI and BHS standards for a device
measuring systolic and diastolic blood pressures
AAMI American Association for the Advancement of Medical Instrumentation
BHS
British Hypertension Society
BP
blood pressure, band-pass (filtering)
CMOS complementary metal-oxide-semiconductor
DC
direct current, zero frequency
DIAS diastolic blood pressure
ECG
electrocardiogram
EMFi trademark for electromechanical film of EMFit, Ltd.
EP
electronic palpation blood pressure measurement method
ESD
electro static discharge
ESH
European Society of Hypertension
HF
high frequency
HW
hardware
IA
intra-arterial
IABP intra-arterial blood pressure
ISH
International Society of Hypertension
Kapton trademark of Dupont, Ltd., for a polyimide material used for PCB and films
LF
low frequency
LP
low-pass (filtering)
MAP mean arterial pressure
OOK on-off-keying
OSC
oscillometric blood pressure measurement method
PAT
pulse arrival time
PC
personal computer
PCB
printed circuit board
PPG
photoplethysmography
PTT
pulse transit time
PWV pulse wave velocity
RAM random access memory
ROM read only memory
SW
software
SYS
systolic blood pressure
™
trade mark
VLF
very low frequency
WHO World Health Organization
List of original papers
I
Ruha, A, Miettinen J, Sorvoja H & Nissilä S (1996) Heart rate measurement based
on detection of the radial artery pressure pulse. Proc. BIOSIGNAL’96, Brno, Czech
Republic, 1: 198–200.
II
Sorvoja H, Kokko V-M, Myllylä R & Miettinen J (2005) Use of EMFi as a blood
pressure pulse transducer, IEEE Transactions on Instrumentation and Measurement
54(6): 2505–2512.
III Nissilä S, Sorvisto M, Sorvoja H, Vieri-Gashi E & Myllylä R (1998) Noninvasive
blood pressure measurement based on the electronic palpation method. Proc. 20th
Annual International Conference of the IEEE Engineering in Medicine and Biology
Society, Chicago, USA, 20(4): 1723–1726.
IV Sorvoja H, Myllylä R, Nissilä S, Kärjä-Koskenkari P, Koskenkari J, Lilja M &
Kesäniemi YA (2001) A method to determine diastolic blood pressure based on
pressure pulse propagation in the electronic palpation method, Proc. 23rd Annual
International Conference of the IEEE Engineering in Medicine and Biology Society,
Istanbul, Turkey, 4: 3255–3258.
V
Sorvoja H, Myllylä R, Nissilä S, Kärjä-Koskenkari P, Koskenkari J, Lilja M &
Kesäniemi YA (2001) Systolic blood pressure accuracy enhancement in the
electronic palpation method using pulse waveform, Proc. 23rd Annual International
Conference of the IEEE Engineering in Medicine and Biology Society, Istanbul,
Turkey, 1: 222–225.
VI Sorvoja H, Hast J, Myllylä R, Kärjä-Koskenkari P, Nissilä S & Sorvisto M (2003)
Blood pressure measurement method using pulse-transit-time. Molecular and
Quantum Acoustics 24: 169–181.
VII Sorvoja H & Myllylä R (2004) Accuracy of the electronic palpation blood pressure
measurement method versus the intra-arterial method. Technology and Health Care,
International Journal of Health Care Engineering 12(2): 145–146.
VIII Sorvoja H, Myllylä R, Kärjä-Koskenkari P, Koskenkari J, Lilja M & Kesäniemi YA
(2005) Accuracy comparison of oscillometric and electronic palpation blood
pressure measuring methods using intra-arterial method as a reference. Molecular
and Quantum Acoustics 26: 235–260.
IX Hast J, Myllylä R, Sorvoja H & Miettinen J (2002) Arterial pulse wave shape
measurement using self-mixing effect in a diode laser, Quantum Electronics 32(11):
975–980.
In the following presentation, the papers above will be referred to by the Roman
numerals I – IX.
Paper I deals with research which employed an eight-channel EMFi sensor to measure
heart rate in non-moving and moderately moving subjects. Both the sensor and the
amplifiers were designed and built by the author, who was also responsible for the test
measurements.
Paper II concentrated on the EMFi sensor material, especially on the temperature
dependence of its sensitivity. Other main concerns were static and dynamic pressure and
statistical sensitivity deviations between the different channels and different sensors. This
work was done and presented earlier in the author’s Licentiate Thesis (1998) and in the
Master’s Thesis (1999) of Vuokko-Marjut Kokko, who conducted her research under the
current author’s supervision. The sensor was further developed and patented by Polar
Electro Ltd. (1998), which also funded the research projects focusing on blood pressure
and pulse measurements during 1993-2000.
The same sensor type was also used in blood pressure measurements. In 1994, the
author demonstrated with self-made EMFI sensors and measurement electronics, which
included a personal computer with data acquisition boards and analysis software that
when cuff pressure on the upper arm drops below the systolic level, the amplitude of
pressure pulses detected in the radial artery increases. Continuing to increase until the
diastolic blood pressure level is reached, the amplitude then stabilizes to a certain level.
The author also developed algorithms that enable determining systolic and diastolic blood
pressure from the amplitude curve. Based on line fittings, these algorithms work both for
decreasing and increasing cuff pressure, as demonstrated by measurements on healthy
volunteers. After that, further development was done by Mika Sorvisto in his Master’s
thesis of 1996. Sorvisto conducted measurements at the Oulu University Hospital on 23
persons, mostly young, healthy volunteers. The author contributed to the research by
developing an eight-channel sensor with amplifiers and band-pass filtering electronics.
The results were published in Paper III. This blood pressure measurement method was
then patented by Polar Electro Ltd. (1999).
It was established in these measurements that there was a difference in the accuracy of
diastolic blood pressure determination between healthy persons and hospital patients.
Therefore, during the winter of 1998/1999, the current author used new measurement
devices to run a second set of measurements on patients who had undergone a cardiac
operation. The method reported in Paper III was based on detecting pulse amplitude
changes, which later proved an unreliable method, leading to the development of another
method, based on pulse transit time changes, as documented in Paper IV. All the
algorithms and analyses in Papers IV-VIII were made by the author. These papers focused
on the accuracy of the developed blood pressure measurement method, which was
compared to measurements performed simultaneously with the intra-arterial and the
oscillometric method.
The last paper (IX) describes an alternative to the use of a pressure sensor to detect
pulsations in the radial artery, namely, the laser-Doppler method. This is a feasible
alternative when the test subject is immobile. In this work, the author concentrates mainly
on providing application-specific advice.
Contents
Abstract
Acknowledgements
List of terms, symbols and abbreviations
List of original papers
Contents
1 Introduction ................................................................................................................... 13
1.1 Motivation to this work ..........................................................................................13
1.2 Contribution of the thesis .......................................................................................14
1.3 Organization of the thesis .......................................................................................15
2 Noninvasive blood pressure measurement methods ...................................................... 16
2.1 Counterpressure with sphygmomanometer ............................................................16
2.2 Auscultatory method...............................................................................................17
2.3 Oscillometric method .............................................................................................19
2.4 Volume clamp, or vascular unloading, method.......................................................22
2.5 Tonometric method.................................................................................................24
2.6 Pressure pulse transit-time method.........................................................................26
2.6.1 Photoplethysmography for the pulse transit time method ...............................28
2.6.2 Other measurement methods and sensors for
the pulse transit time method...........................................................................30
2.6.3 Standards and accuracy validation...................................................................31
2.7 Blood pressure values.............................................................................................32
2.7.1 Clinically approved blood pressure measurement devices ..............................33
3 EMFi pressure sensor array ........................................................................................... 34
3.1 EMFi sensor material..............................................................................................34
3.2 Preparation of blood pressure sensor ......................................................................36
3.3 Measurement setups for the blood pressure sensor.................................................38
3.4 Sensor measurements .............................................................................................41
4 Blood pressure measurements ....................................................................................... 48
4.1 Measurement setups ...............................................................................................48
4.2 Study subjects .........................................................................................................49
4.3 Blood pressure determination in the electronic palpation method..........................51
4.4 Accuracy of the electronic palpation method..........................................................58
4.5 Blood pressure determination in the oscillometric method ....................................61
4.6 Accuracy of the oscillometric method ....................................................................65
5 Discussion ..................................................................................................................... 67
6 Conclusion..................................................................................................................... 72
References
Original papers
1 Introduction
1.1 Motivation to this work
Cardiovascular diseases are one of the leading causes of death. Globally, they underlie
the death of one third of the world’s population and among Europeans, such as the
Finnish, one half of all deaths are attributed to them. These diseases can be divided into
coronary, cerebral or peripheral artery diseases. The main cause of cardiovascular
diseases is atherosclerosis, in which arterial compliance decreases dramatically with the
result that arterial stiffness increases (van Popele et al. 2000 and 2001), which serve to
increase the patient’s death risk (Mattace-Raso et al. 2006). This ties in closely with
hypertension and type 2 diabetes. Noninvasive measurements for assessing these diseases
would be highly valuable for obvious reasons.
Such invasive blood pressure measurement methods as are generally used in hospitals,
especially in intensive care units, are known to carry a risk, albeit a small one. Risks of
these kinds could be avoided, however, if there was a noninvasive method offering a high
degree of accuracy and real time operation in a continuous, beat-to-beat, mode. Further,
the method should be insensitive to the patient’s movement (artifacts) and respond
rapidly to cardiovascular changes, such as a sudden drop in blood pressure. As it is,
invasive methods are currently prevalent.
In clinics and home care, noninvasive blood pressure measurement devices have
become increasingly common during the last decade as their prices have sunk to an
appropriate level for ordinary consumers. Automatic measurement features and easiness
of use have also contributed to their growing popularity. Nevertheless, the accuracy of
these devices has not yet reached the necessary level, since only some of them are
clinically validated and most have a questionable accuracy (O’Brien 2001).
Originally, the stimulus that led to the work presented in this thesis came from
Professor Seppo Säynäjäkangas, the founder of Polar Electro Ltd., a long-time market
leader in ambulatory heart rate monitors. The idea was to design a small-sized,
ambulatory, low-power noninvasive measurement device and to develop a measurement
method for sensing blood pressure pulsations in the radial artery. These pulsations can be
used to determine heart rate and blood pressure which, in addition to ECG
14
(electrocardiography) are undoubtedly the most important and most commonly measured
biosignals.
An ideal site for the measurement is provided by the radial artery in the wrist, since it
is located near the skin surface and the wrist serves as a good attachment for the
measuring device. In addition, both patients and healthy consumers are unlikely to object
to this place of measurement. Moreover, measuring the shape of the arterial pressure
pulse and wave velocity allows additional cardiovascular parameters to be assessed,
including arterial stiffness, which is a manifestation of atherosclerosis. During the past
decade, as measurement methods and sensor technologies have advanced, pulse wave
velocity (PWV) measurement has become broadly used and clinically approved (Asmar
et al. (1995) and Asmar (1999)). Low-power radial artery pressure sensors are attractive
to use, because, together with other signals such as the ECG, they can be used for PWV
measurements in addition to heart rate and blood pressure measurements.
1.2 Contribution of the thesis
Only a few sensing methods and materials, powered by a single button cell battery, are
applicable to long-term, ambulatory monitoring of blood pressure pulses, although these
requirements need to be fulfilled in sporting and home care applications as well as in
clinical use and intensive care. Of these, sporting applications in particular call for a
wider temperature range than the others, which are usually carried out at room
temperature. Ideally suited for measurements at that temperature is the recently
developed sensor material known as electromechanical film (EMFi).
The main contribution of this thesis is the development of a new method for
noninvasive blood pressure pulse monitoring in the wrist, based on using EMFi material
to obtain systolic and diastolic blood pressures. To determine the presented method’s
accuracy, the intra-arterial, invasive method was used as a reference. Next, these values
were compared to those obtained by the oscillometric method, which is the most common
commercially available method. All measurements using the three different methods were
conducted simultaneously. This work could be divided into seven steps:
1. Reviewing current noninvasive methods of determining blood pressure, comparing
their accuracy rates and explaining their historical backgrounds.
2. Developing multi-channel pressure transducer arrays that are accurate enough for
blood pressure pulse detection and blood pressure determination in the wrist.
3. Developing a measurement system for clinical monitoring and data acquisition.
4. Making test measurements with patients in the hospital environment.
5. Conducting experiments to noninvasively determine diastolic blood pressure in the
wrist based on pulse amplitude and pulse transit time, and assessing possible factors
affecting the achieved measurement accuracy. This step also involved carrying out
experiments to noninvasively determine systolic blood pressure in the wrist and using
pulse shapes to enhance measurement accuracy.
6. Comparing the accuracy of the presented method to that of the oscillometric method.
7. Summarizing the results and assessing the feasibility of the solution.
15
1.3 Organization of the thesis
This thesis consists of six chapters. Following Introduction, Chapter 2 presents a
description of widely available noninvasive methods of measuring blood pressure,
complemented by a detailed account of them and their historical backgrounds. At the end
of the chapter, a presentation is given on the classification of noninvasive blood pressure
measurement devices, their accuracy requirements and methods of determining accuracy.
The next chapter describes the design, construction and testing of low-power blood
pressure pulse sensors and their use in the detection of blood pressure pulses and
noninvasive determination of blood pressure.
Next comes a description of the measurements system, comprising an automated arm
cuff and a wrist sensor as well as a data acquisition and conditioning system that is
connected to a patient monitoring unit and a personal computer. Experimental
measurements were conducted at the Oulu University Hospital among young, healthy
volunteers and elderly cardiac patients, whose characteristic blood pressure values are
presented at the end of the chapter. These measurements were conducted simultaneously
with two noninvasive measurements using the oscillometric and the new electronic
palpation method introduced here. The chapter goes on to show the measurement results
and assess the methods’ accuracies using intra-arterial blood pressure measurements as a
reference. Chapter 4 presents detailed findings for the methods.
Chapter 5 discusses the significance of the results, while Chapter 6 provides a
summary of this work.
2 Noninvasive blood pressure measurement methods
A historical overview of the different noninvasive blood pressure measurement methods
will be given here together with a description of each. Most of the historical data are
gleaned from Geddes’ Handbook of Blood Pressure Measurement (1991), complemented
by new information put out after its publication. Particular emphasis will be placed on
detailing the methods’ strong and weak points and on the achieved accuracies. Towards
the end, I shall show the accuracy of a number of blood pressure measurement devices.
The pressure unit used systematically in this thesis is mmHg, which, although not an SI
unit, is nonetheless a globally approved blood pressure unit for historical reasons.
2.1 Counterpressure with sphygmomanometer
Noninvasive blood pressure measurement methods are indirect and based mainly on
measuring counterpressure. Putting a hand in a water-filled chamber, and sealing it with
the wrist, allows the measurement of pressure alterations in the chamber. This pressure
alteration, caused by blood pressure oscillations in the hand, depends on the static
pressure within the chamber. If this pressure is increased, the pressure alteration will rise,
reach its maximum, and then decrease until blood circulation in the hand stops altogether.
In 1876, Etienne Jules Marey, the French physiologist, first used this method for
noninvasive blood pressure assessment in humans.
In the same year, Von Bash employed a water-filled bag that was connected to a
mercury manometer. When this bag was pressed with increasing force against the skin
over the radial artery, pulsations could be observed until, after a certain point, they
disappeared. At this point, counterpressure was assumed to represent systolic pressure.
Hydraulic counterpressure was subsequently replaced by pneumatic pressure. Marey
designed an instrument, consisting of a rubber sleeve in a rigid cylinder, a mercury
manometer and a pump. Once a finger was attached into the cylinder, weak pulsations
could be observed in the mercury level. Potain improved this measurement system in
1902, by replacing the mercury manometer with a gauge and substituting water for air.
In 1895, Mosso improved Marey’s measurement system by applying counterpressure
on all fingers, thereby obtaining a larger signal.
17
Almost at the same time, Scipione Riva-Rocci (1896) in Italy and Hill and Barnard
(1897) in England developed a cuff, which was attached around the upper arm, thus
occluding the brachial artery when pressurized. In both these methods, the used cuffs
were too narrow according to Von Recklinghausen (1901).
In 1899, Gärtner introduced a method to determine systolic blood pressure by placing
a cuff around a finger, pressing blood off from the distal side of the cuff, inflating it to a
point above presumed systolic pressure, and then deflating cuff pressure. When cuff
pressure reached the systolic blood pressure level, blood flushed to the finger tip.
Gärtner’s method has since been known as the blush or flush method.
As the pneumatic sleeve or cuff, also referred to as a sphygmomanometer, consisted
usually of a rubber bladder woven inside a fabric, it enabled indirect, noninvasive arterial
blood pressure measurements. Up until the present day, the cuff has been the single most
important component of noninvasive blood pressure measurement devices. The cuff is
usually attached around the arm on the brachial artery or, in some devices, around the
wrist or even a single finger. Pressure measurements can be made by using a mercury or
aneroid manometer, which have become the “gold standards” in clinical use, but are now
being replaced with silicon-based pressure sensors inside automatic measurement
devices, based usually on the auscultatory or oscillometric method.
2.2 Auscultatory method
In his MD thesis of 1905, Nikolai Korotkoff, from St. Petersburg, Russia, presented blood
pressure measurements performed on animals using a Riva-Rocci sleeve, mercury
manometer and child stethoscope. First, he quickly raised cuff pressure until it stopped
blood circulation on the distal side of the hand, indicated by palpating the radial artery.
During the following slow pressure drop, audible sounds could be heard through the
stethoscope, which was placed on the skin beyond the sleeve. These sounds were affected
by the blood wave in the artery under the cuff, and were audible at 10-12 mmHg, slightly
before the pulse could be palpated on the radial artery. At this point, cuff pressure is taken
to indicate maximum blood pressure, while minimum blood pressure is achieved when
the murmur sounds disappear. (Geddes 1991)
Korotkoff sounds were corroborated by the British researchers MacWilliam and
Melvin (1914) and the American Warfield (1912), who used intra-arterial pressure
measured from a dog as a reference. Then in 1932, Wold and von Bonsdorff applied
Korotkoff’s auscultatory method on humans using intra-arterial blood pressure as a
reference.
Later, the characteristic sounds heard during cuff deflation were divided into five
phases on the basis of their intensity. First to publish were Goodman and Howell (1911),
followed by Grödel and Miller (1943), Korns (1926) as well as Rappaport and Luisada
(1944).
The appearance and disappearance of sound can be used to determine systolic and
diastolic blood pressure, respectively, while the cuff deflates. However, establishing the
point at which sounds disappear is not always obvious; in fact, misleading readings are
easy to record. Thus, a certain sound intensity level is often used to determine the point
18
corresponding to diastolic blood pressure. Also the sound amplitude maximum can be
used to determine mean arterial blood pressure, as demonstrated by Davis and Geddes
(1989 and 1990).
The origin of the sounds has been discussed since Korotkoff’s initial findings.
McCutcheon and Rushmer have provided the latest universally acknowledged
explanation (1967): “An acceleration transient produced by an abrupt distension in the
arterial wall as a jet of blood surges under the cuff into the distal side of the artery. This
produces the first tapping sound which signals systolic pressure. It continues as cuff
pressure is decreased and disappears at diastolic pressure.” Also, they mentioned that
“Turbulent or eddy flow, which follows the initial jet, and produces audible sound. This
factor plays little or no significance in the auscultatory technique.”
It should be pointed out that the genesis of the sound will be different in the four
phases; one mechanism predominating in one phase and a different one in another.
Geddes (1991) summarised in his Handbook of Blood Pressure Measurement that other
factors can also be pointed out, such as the rate of the pressure increase, which has a
direct effect on sound intensity (Ehret 1909, Bramwell 1926 and Tavel 1969). As a result,
the auscultatory method may yield erroneous results in hypertension (Rodbard 1962,
1967 Pederson and Vogt 1973). In addition, fluid velocity plays an important role, as
explained by Flack (1915). Thus, a sufficient flow velocity is a precondition for the
audibility of Korotkoff sounds (Rappaport 1944, Rodbard 1953, and McCutcheon 1967).
Measurements based on the auscultatory method are difficult to automate, because the
frequency spectrum of the different phases of Korotkoff sounds is closely related to blood
pressure. When a patient’s blood pressure is high, also the recorded frequency spectrum is
higher than normal and decreases as a function of blood pressure. With hypotensive
patients and infants, on the other hand, the highest spectrum components can be as low as
8 Hz (Whitcher et al. 1966 and 1967), which is below the human hearing bandwidth.
Normotensive subjects, in turn, require a bandwidth of 20 Hz to 300 Hz for a sufficient
reproduction of Korotkoff sounds (Geddes 1991), although most of the energy of the
signal spectrum is below 100 Hz.
Since technology improvements have enabled efficient signal processing and
calculation, a number of papers have been published on the classification of Korotkoff
sounds. Cozby and Adhami (1993), for example, discovered the sub-audible, low
frequency part of Korotkoff sounds and concluded that energy in the 1-10 Hz bandwidth
of the total energy rises from 60% to 90%, when cuff pressure decreases from above the
systolic to a level below it. This feature can be used as a thresholding algorithm for
determining systolic blood pressure. Regueiro-Gómez and Pallás-Areny (1998) were able
to use the spectral energy dispersion ratio to determine systolic and diastolic blood
pressure with high precision: in 97% of all cases, the determined values for 15 persons
were within ±1 heart beat.
In ambulatory measurements, when the patient is able to move moderately, noise may
become dominant, thereby spoiling the measurement. This may be avoided by using two
identical microphones under the cuff, one located on the upper side, the other on the
distal side. Ambient noise reaches both microphones at the same time, but the blood
pressure pulse propagating through the brachial artery arrives after a time delay. This
phenomenon can be used for noise cancellation as described by Sebald et al. (2002).
19
TM
Allen et al. (2004) used a Matlab -based short-time Fourier transform joint timefrequency analysis to obtain spectral information of Korotkoff sounds. Their results show
that phase III has the largest overall amplitude and high frequency energy. In contrast,
phases IV and V have the lowest amplitude and frequency components. Statistically
significant transitions were observed between the different phases; thus, the transition
from phase I to II was characterized by increases in high frequency and sound duration,
the transition from phase II to III by a significant decrease in sound duration, the
transition from phase III to IV by decreases in maximum amplitude, highest frequency
and relative high frequency energy, and the transition from phase IV to V by decreases in
the maximum amplitude and high frequency energy. They concluded that the method was
able to detect the five phases also identified by an expert cardiologist, verifying that the
method can be used in automatic blood pressure measurements.
2.3 Oscillometric method
As discussed previously, the oscillometric method was first introduced by Marey in 1876.
Erlanger (1904) further developed it and attached a Riva-Rocci cuff around the upper arm
instead of fingers. Pressure oscillations were recorded on a rotating drum. This graphic
recording was discarded by Pachon (1909), who used dual-dial gauges, one for presenting
oscillation amplitude and the other for showing cuff pressure.
In those days, it was thought that maximum oscillation amplitude indicated diastolic
blood pressure (Howell and Brush 1901). It was more than half a century later that Posey
and Geddes showed that that point actually corresponds to mean arterial pressure.
Ramsey (1979) and Yelderman and Ream (1979) verified the finding with adults, while
Kimble et al. did the same with newborn infants (1981). At around the same time,
Alexander et al. (1977) demonstrated that the cuff width should be at least 40% of the
arm circumference.
Diastolic and systolic blood pressures can be determined using special fractions of the
maximum oscillation amplitude, also known as characteristic ratios. Thus, the special
fractions for systolic and diastolic blood pressure are 50% and 80%, respectively. Friesen
and Lichter (1981) used these characteristic ratios to determine the diastolic and systolic
blood pressures of infants, neonatal and pre-term babies with excellent results: the
regression line formula was D = 0.94p + 3.53 for systolic pressure and D = 0.98p + 1.7
for diastolic pressure.
Geddes et al. (1982) made measurements with animals (13 dogs) using the direct
invasive method as a reference and with humans (43 adult subjects) using the
auscultatory method as a reference. Auscultatory signals inside the cuff were recorded
using a tiny piezoelectric microphone with a bandwidth of 30-300 Hz to record Korotkoff
sounds and another with a bandwidth of 0.3-30 Hz to obtain cuff oscillations, both
recorded simultaneously. For humans, the characteristic ratios were 45%-57% and 75%86% for systolic and diastolic pressure, respectively. For animals, these values varied
between 43%-73% and 69%-83%. Inevitably, these ratios vary considerably and more
theoretical studies are needed.
20
Drzewiecki et al. (1994) used a mathematical model to study the theory that the
oscillation maximum appears at mean arterial pressure. Their model supported the
hypothesis. In addition, they found the average characteristic detection ratios to be 59%
for systolic and 72% for diastolic pressure. They also suggest that the systolic ratio
should be lower for hypertensive patients and that the diastolic ratio should be lower for
hypotensive patients.
Cristalli et al. also conducted experimental studies (1992-1994) using a pneumatic
system. Later, they constructed mathematical models to describe pressure distribution and
the influence it has on tissue properties under the cuff. With a computer-controlled
system, they demonstrated that frequent cuff inflations and deflations may significantly
decrease the recorded diastolic and systolic blood pressure values. With some patients,
the decrease, brought about chiefly by a decrease in Poisson’s ratio, may exceed 10
mmHg. Ursino and Cristalli (1994 and 1996) constructed a lumped parameter
mathematical model which took into account compliance of the occluding cuff, pressure
transmission from the cuff to the brachial artery through soft arm tissue and blood
changes occurring in the collapsing brachial artery under the external load. In addition,
they showed that the rigidity of the arterial wall and tissue compliance have a significant
effect on the accuracy of diastolic and systolic pressure values, causing an error of 15%20%. By contrast, changes in mean arterial pressure and cuff compliance did not
influence the measurement to a large degree. They also supposed that the mean arterial
blood pressure can be determined as the lowest pressure at which cuff pulse amplitude
reaches the plateau. Another finding by Ursino and Cristalli was that the characteristic
ratios are also greatly affected by the arterial wall’s elasticity; thus, excessively elastic
arteries may lower the systolic ratio by 25%-30% and this figure may rise to 80% with
stiffened arteries. According to their model, arterial stiffness has a considerably weaker
effect on the diastolic ratio. Typical characteristic ratio values were 46%-64% for the
systolic ratio and 59%-82% for the diastolic ratio. As a conclusion, a measurement device
with fixed ratios for determining systolic, mean and diastolic blood pressure may
significantly overestimate them.
Moraes and Cerulli (1999 and 2000) also studied the characteristic ratios using
computer-controlled linear cuff pressure deflation with 10 patients and 75 volunteers,
using the auscultatory method as a reference. A fixed percentile rule yielded a value of
56% for systolic and 76% for diastolic pressure. With these ratios, they got an average
errors and standard deviations of error of (-0.9 ± 7.0) mmHg and (1.0 ± 6.5) mmHg for
systolic and diastolic blood pressure, respectively. This format (average ± standard
deviation) will be used anytime in the text if not specified in other terms. Applying an
adaptive classification rule, they changed the ratios in conformity with the arm
circumference and mean blood pressure, i.e., the systolic ratio was decreased from 64%
to 29%, mean artery pressure was changed from less than 70 mmHg to over 150 mmHg
and diastolic pressure was increased from 50% to 75%. The obtained accuracies were (1.5 ± 5.1) mmHg for systolic and (0.6 ± 5.9) mmHg for diastolic pressure. As can be
seen, the standard deviation decreased slightly. In general, the deviations in the
characteristic ratios recorded by Moraes and Cerulli ranged widely.
Being dependent on the used ratios, the fixed percentile rule will lead to erroneous
readings in many measurements. In addition, motion artefacts, the white coat effect,
shivering and arrhythmia generate noise that has an adverse effect on the accuracy of the
21
readings. Lin et al. showed in a paper published in 2003 that, to minimize noise, different
mathematical methods can be used, including averaging, also known as smoothing,
Kalman filtering or fuzzy logic discrimination with recursive weighted regression. Colak
and Isik (2003 a & b and 2004 a, b & c) utilized neural networks and fuzzy logic to
classify blood pressure profiles. Because most manufacturers of automatic devices do not
reveal exactly how their device determines the pressure values, the only option available
to researchers is to test them with a large number of patients with different conditions,
including hypo-, normo- and hypertension, arterial stiffening, heart disease, arrhythmia
and so forth.
Bur et al. (2000) studied the accuracy of one commercial (Hewlet Packard)
oscillometric blood pressure monitor. They tested three different cuff widths on their
subjects, who were critically ill patients, using intra-arterial (radial artery) pressure as a
reference. The authors concluded that the device significantly underestimated arterial
blood pressure and produced a high number of measurements out of the clinically
acceptable range. Therefore, this type of oscillometric blood pressure measurement fails
to achieve an adequate accuracy level with critically ill patients.
Reeben and Epler (1973 and 1983) were the first to suggest using a modified
oscillometric measurement system, where cuff pressure changes beat-to-beat on the basis
of the actual mean pressure in the artery. In 1996, Jagomägi et al. presented a continuous
differential oscillometric blood pressure method, which was capable of continuously
measuring mean arterial pressure. The device (UT9201) used two servo-controlled cuffs
around two adjacent fingers, one for mean pressure and another for slowly changing the
pressure between a value higher and lower than the mean. Jagomägi’s measurements
were carried out on young volunteers using the volume clamp method (Finapres) as a
reference. In many of the measurements, variations in mean arterial pressure were very
similar. The achieved accuracy for mean blood pressure was (-1.1 ± 5.5) mmHg at rest,
(0.5 ± 6.9) mmHg in head-up tilt and (-3.6 ± 7.7) mmHg during deep breathing. Raamat
et al. (1999) studied maximum oscillation criteria and found that, with contracted finger
arteries, the method gives overestimated values of up to 19 mmHg. With relaxed arteries,
however, the error is less evident. In their next paper (2000a & b), Raamat et al. tested
accuracy during local hand heating and found no statistical error in the absence of
vasoconstriction ((0.3 ± 0.3) mmHg). With peripheral vasoconstriction, on the other hand,
there was a statistically significant difference ((6.7 ± 2.0) mmHg). Raamat et al. (2001)
also studied whether accuracy is affected by cooling the arm locally; their results were
statistically significant ((-1.5 ± 1.1) mmHg before cooling and (8.8 ± 6.3) mmHg after
cooling). Jagomägi (2001) recommended that intensive vasoconstriction, which can be
determined by a laser-Doppler skin flow meter, should be avoided during measurements.
Further, Raamat et al. (2001) conducted a set of measurements during a rhythmical
quadriceps exercise and simultaneously performed static handgrip compression using a
rapidly (0.2 s) inflating and deflating occlusive cuff in the right hand. Finapres and
UT9201 cuffs were fixed around the subjects’ left hand fingers. The aim was to affect
peripheral vascular resistance and blood pressure, and the results showed that both
devices tracked mean arterial pressure similarly. The test groups’ averaged difference was
1.2 mmHg after successive inflations and 3.8 mmHg after successive deflations.
22
2.4 Volume clamp, or vascular unloading, method
Penaz invented the continuous noninvasive finger blood pressure measuring method
based on a transparent finger cuff, photoelectric plethysmogram, i.e. transmitter and
detector, and pressure controller unit in 1973. Molhoek et al. (1984) then tested it with 21
hypertensive patients using intra-arterial pressure as a reference. They obtained an
underestimation of 6 mmHg for both systolic and diastolic blood pressure, making the
method attractive for further analysis. Gravenstein et al. (1985) studied tissue hypoxia in
order to assess its impairing effect on blood circulation in the finger tip on the distal side
of the finger cuff. They observed an oxygen desaturation from 97% to 93.7%, recorded
after applying cuff pressure for one minute. Boehmer (1987) from Ohmeda Monitoring
Systems introduced a prototype device, later known as FinapresTM, based on the vascular
unloading method, in which the cuff on the finger is pressurized to the same pressure as
the artery. Cuff pressure was controlled by electronic, automatic adjustments so that it
equalled the intra-arterial pressure at all times and pulsated with arterial pressure. These
measurements consisted of photoplethysmographic measurements of finger volume; a
light source and a detector located inside a transparent cuff measured blood volume
alterations, i.e., blood flow. Cuff pressure was controlled by servo technique using a
microprocessor to keep the blood flow stable; when the pressure inside the artery
increased, external cuff pressure increased by the same amount at the same time. The
system automatically corrected changes induced by smooth muscle contractions or
relaxations.
Factors affecting Finapres’ reliability and optimal measurement conditions were
studied by Kurki et al. (1987). They measured 50 men during surgical operations using
four different fingers and three different types of cuff and established that, compared to
the intra-arterial blood pressure, the highest accuracy was achieved with the thumb: the
observed accuracies were -4.8 mmHg for systolic, 1.49 mmHg for diastolic and 0.29
mmHg for mean arterial pressure. The correlation coefficients were 0.945, 0.884 and
0.949, respectively, indicating high accuracy. Epstain et al. (1989), however, studied
obstetric patients during spinal anaesthesia and compared Finapres’ accuracy with that of
other automated oscillometric methods. Differing greatly from previous studies, their
results indicated an accuracy of (6.6 ± 12.5) mmHg for systolic, (3.3 ± 10.4) mmHg for
mean and (7.2 ± 9.8) mmHg for diastolic pressure. Later Kugler et al. (1997) ran a series
of tests with a sphygmomanometer and concluded that there were systematic differences
in the results of the different methods. It must be pointed out, however, that their
reference method was noninvasive and thus less accurate than the invasive method used
by Kurki et al. Shortly after, Kurki et al. (1989) published a study on open-heart surgery,
which demonstrated that Finapres worked well with 13 patients out of 15, before they had
a cardiopulmonary bypass operation and in 10 out of 15 during the operation. They
reported average accuracies of (8.3 ± 10.2) mmHg and (6.6 ± 8.7) mmHg with the
correlation coefficients of 0.814 and 0.902, respectively. Since then, a number of papers
have explored the accuracy of Finapres. Bos et al. (1996) and Wesseling (1996) have also
defined a specific correction method for Finapres and Portapres (Shmidt 1991), in which
the arterial pressure waveform recorded on a finger is filtered and bias-corrected to
closely correspond to intra-brachial pressure. To conclude this discussion, it must be
23
pointed out that, as a result of extensive combined studies conducted by Silke &
McAuley (1998) and Imholz et al. (1998), the accuracy of Finapres is now broadly
acknowledged. More than 150 papers have been published on blood pressure
measurements using Finapres, some of which are tabulated in Table I.
Table 1. Comparison of the accuracy (mean values ± standard deviations) of the
Finapres (also Portapres and Finometer) and intra-arterial pressure (IAP).
Year
Authors
1991
Schiller Z &
Reference
method
IAP
I: 15 Two groups,
II:16 satisfactory reliable
IAP
Not reliable during
Pash T
1991
Epstein R H
Num. Comments
et al.
1992
Bos W J
Systolic
Mean
Diastolic
accuracy
accuracy
accuracy
(mmHg)
(mmHg)
(mmHg)
I:-1.6 ± 16.6, I: 0.2 ± 11
I: 4.6 ± 11.5
II:-2.7 ± 13.1 II:-2.6 ± 10.9 II: 4.7 ± 9.8
5.8 ± 11.9
7.7 ± 10.0
8.2 ± 9.8
-15.7 ± 18.8
-20.1 ± 15.7
-13.5 ± 15.7
anaesthesia
IAP
12
et al.
Hypertensive and
vascular disease
1994
Novak V et al.
IAP
20
One hour recording
0.84 ± 13.3
6.67 ± 5.23
1994
Triedman J K
IAP
27
Operated pediatrics
-18.6
-13.4
IAP
15
Elderly at rest
-16.8 ± 2.6
-10.8 ± 1.5
-17.5 ± 1.6
IAP
brachial
28
Healthy elderly and
-3.2 ± 16.9
-3.7 ± 7.0*
-13.0 ± 10.5
-0.7 ± 4.6*
-8.4 ± 9.0
1.0 ± 4.9*
et al.
1995
Rongen G A
et al.
1996
Bos W J
et al.
1998
Silke B &
vascular diseased
IAP
patients
449 Combined databases
2.2 ± 12.4
2.1 ± 8.6
0.3 ± 7.9
IAP
1031 Combined databases
-0.8 ± 11.9
-1.6 ± 7.6
-1.6 ± 8.3
1 ± 10
1 ± 11*
-10 ± 8
-2 ± 7*
-8 ± 7
-2 ± 7*
-9.7 ± 13.0
-16.3 ± 7.9
-11.6 ± 8.0
-1.1 ± 10.7*
3.1 ± 7.6*
-1.5 ± 6.6*
2.7 ± 4.7*
-0.2 ± 6.8*
4.0 ± 5.6*
McAuley D
1998
Imholz B P M
et al.
2002
Westehof B E
et al.
2003
Guelen I et al.
IAP
brachial
14
IAP
brachial
37
Hypertensive, 24 h,
Portapres, correction
Age 41-83
Finometer,
1st correction
nd
2 correction
*Corrected
Finapres, Portapres and Finometer are widely used in different types of physiological
measurements. For example, in 60 degree head-up tilt tests (Friedman 1990), the obtained
blood pressure variation correlation coefficient was 0.98 with intra-arterial blood
pressure. Hildebrant et al. (1991) and Penzel (1993) have performed a range of studies on
exercise and sleep.
Demonstrating that the volume clamp method can also be used in other ways,
Sasayama et al. presented in 1994 findings from studies involving a cuff and a
photoplethysmogram attached on the ear. The average errors were 3.7 and 2.7 mmHg for
systolic and diastolic pressure, respectively.
24
Finapres devices are no longer available, since they have been replaced by another
device, known as Finometer, a trademark of FMS, Finapres Medical Systems BV,
Arnhem, Netherlands. The company also produces PortapresTM blood pressure monitors,
which measure two fingers alternately, enabling one finger to normalize while the other is
being measured. Langewouters et al. (1998) encourage physicians to use Finapres or
Portapres instead of invasive methods.
Currently, the latest journal paper on the topic was published by Maestri et al. in 2005.
With Finometer and Finapres, they conducted continuous blood pressure measurements
lasting approximately four hours on 19 cardiac patients (including patients with
myocardial infarction) using intra-arterial blood pressure as a reference. Systolic and
diastolic values were automatically defined for each signal. They divided the bandwidth
into three bands, VLF (0.01-0.04 Hz), LF (0.04-0.15 Hz), HF (0.15-0.45 Hz) and
calculated the power spectral density for each band. In addition, percentage errors were
computed for all these bands and coherence and gain functions were obtained. With both
devices, systolic and diastolic signals showed a high rate of errors in the VLF and LF
regions. As a result, the authors concluded that the accuracy of Finometer exceeds that of
Finapres in noninvasive measurements of absolute arterial pressure. However, in blood
pressure variability measurements it produces worse results in most spectral parameters.
When autonomic assessment is based on analyzing spontaneous blood pressure
variability of pure cardiovascular origin, it may be preferable to use Finometer’s Finapres
output instead of a reconstructed brachial artery waveform.
2.5 Tonometric method
First presented by Pressman & Newgard in 1963, the arterial tonometer is a pressure
measurement method that can noninvasively and continuously record pressure alterations
in a superficial artery with sufficient bony support, such as the radial artery. It uses a
miniature transducer (e.g., by Millar Instruments) or a rigid sensor array (e.g., by Colin
Electronics) or a flexible diaphragm (Drzewiecki et al. 1983), which is attached on the
skin above the pulsating artery. Skin and tissue located between the sensor and the array
transfer pressure pulsations between them. When the pulsations reach their strongest
level, the sensor is regarded as being correctly positioned. This can be facilitated by using
a sensor array and selecting sensor elements with the strongest amplitude. This method
requires that the sensors are closely alike in terms of sensitivity. Next, the sensor or
sensor array is pushed towards the vessel using, for example, air pressure. The vessel
flattens when the pressure and, consequently, the force against the artery wall increases.
Arterial pressure in the top of the flattened artery’s center equals the supporting pressure,
allowing the recording of an accurate blood pressure profile. If the pressure increases too
much, the artery will occlude totally and the measurement will be erroneous. Thus, the
hold-down pressure must be continuously controlled using stress distribution information.
Furthermore, the size of the sensor or sensor element must be small relative to the artery
and the sensor material must be rigid, which has led to the use of piezoelectric or
piezoresistive (made of silicon) materials. Moreover, sensor arrays enable the use of
motion artifact cancellation methods to improve the signal-to-noise ratio, as in the
25
piezoelectric sensor arrays presented by Ciaccio & Drewiecki in 1988 and Ciaccio et al.
in 1989. The measurement method and the piezoresistive silicon sensor array were
presented by Weaver et al. in 1978 and by Terry et al. in 1990. Developed in the USA in
co-operation with SRI International & IC Sensors, the sensor has been marketed by Colin
Electronics, Co., Ltd., Japan (Ohmori 1992) since 1992. Zorn et al. (1997) validated the
tonometer monitor, known as Colin Pilot 9200, with 20 adult and paediatric patients
using intra-arterial blood pressure as a reference. They achieved an accuracy of (2.24 ±
8.7) mmHg and (0.26 ± 8.88) mmHg, which slightly exceeded the allowable standard
deviation of error. Since then, the Colin tonometer has been used to estimate the
waveform of central aortic blood pressure (Hori et al. 1997, Chen et al. 1997, Fetics
1999) using mathematical transformation, i.e., a transfer function between a noninvasive
tonometer and an invasive catheter measurement site. An autoregressive exogenous
model of the Colin tonometer signal proved accurate in studies by Hori (approximately a
3% estimation error for systolic, mean and diastolic pressure) and Fetics ((0.4 ± 2.9)
mmHg). Chen studied a generalized transfer function model and obtained an almost
identical reading ((0.2 ± 3.8) mmHg). Nowadays, the Colin Pilot is represented by DRE
Medical, Inc., USA.
Another type of tonometer is based on volume changes in a fluid-filled, flexible
diaphragm chamber (Bansal et al. 1994, Drewiecki 1995, Drewiecki et al. 1996). In this
setup, the sensor base is made of plexiglass. A rounded rectangle-shaped channel is
machined into it, and the sensor array is sealed with a thin polyurethane sheet,
functioning as a flexible diaphragm. Next, the channel is filled with saline. At each end of
the channel, there are two stainless steel electrodes used to inject a current along the
channel length. Near the centre of the channel, four measuring electrodes are placed at
equal spacing. Each pair of electrodes defines a volume compartment and the voltage
across each pair is calibrated in terms of volume using impedance plethysmography.
External to the tonometer, a saline-filled catheter is used to connect the channel to an
electro-mechanical volume pump. A Velcro strap can be used to fix the tonometer in
place. When the tonometer is attached on the skin over a superficial artery, such as the
radial artery, blood pressure pulsations in it produce volume shifts inside the volume
sensor resulting in an impedance change. The pressure inside the tonometer can be varied
so that the strongest signal can be measured. This pressure corresponds to mean arterial
pressure, MAP. Liu & Lin (2001), Wang et al. (2002) and Lin et al. (2003) have proposed
a synthetic fuzzy-logic controller with three parallel subcontrollers and a model-based
linear predictor for a similar tonometer (albeit with only four electrodes) designed to
track the tendency of mean arterial pressure in the radial artery. (Drzewiecki 1995)
In 2002, Nelesen and Dimsdale studied the accuracy of the Colin Pilot in
cardiovascular reactivity testing using a brachial artery measurement (Critikon Dinamap)
as a reference. They compared the results with those of Finapres and concluded that the
tonometer offered low artefact rating and high accuracy. In some measurements, Finapres
overestimated the results by more than 40 mmHg and showed a considerable number of
artefacts (75%). On the other hand, the Pilot tracked blood pressure in the same subject
without error.
Recently, in 2004, Salo et al. proposed a monolithically integrated single-channel
capacitive tonometer sensor chip and demonstrated its use. Developed using industrial
26
CMOS technology combined with post-process micromachining, the chip enables the
design of small, portable devices with quite low power consumption (11.5 mW).
2.6 Pressure pulse transit-time method
Galen (131-200 AD) was the first in Europe to note the relationship between the heart
and the arteries, and he also pointed out that the arteries contain blood, not air. He ranks
as the foremost sphygmologist of Antiquity, not merely because he wrote a number of
books on the subject, but because his teaching on the pulse dominated clinical practice
for about 16 centuries. The first attempt to represent the pulse graphically, in a form
comparable with sphygmographic tracing, was made in 1540 by Struthius, who studied
pulse waves by placing a leaf on the radial artery and watching its vibrations. It was only
in 1553 that Miguel Serveto described pulmonary circulation, and in 1628 that William
Harvey provided the first description of blood circulation. Harvey’s work established the
pulse wave as a manifestation of cardiac ejection. More than 100 years later, in 1731, the
Reverend Stephen Halens recorded the first invasive arterial pressure measurement.
Etienne Marey, in turn, was the first to accurately record arterial pulses in humans using a
sphygmograph. Later, several sphygmographic devices were devised. (Asmar 1999)
It was recognized quite early that the elasticity of an artery is related to the velocity of
the volume pulses propagating through it. Moens (1878) and Korteweg (1878) derived a
mathematical expression for the velocity of the pulse front traveling along an artery as a
function of such factors as the elasticity coefficient, the thickness of the arterial wall and
the end-diastolic diameter of the vessel lumen. This is shown in the formula below, where
PWV is pulse wave velocity; t the thickness of the vessel wall, d the diameter of the
vessel, ρ the density of blood and E stands for Young’s modulus describing the elasticity
of the arterial wall (Asmar 1999, McDonald 1974)
PWV=
tE
.
ρd
(1)
Pulse transmission velocity was used as a clinical index of arterial elasticity by Bramwell
and Hill (1922), and later by a large number of authors using different methods. In 1981,
Geddes showed that diastolic blood pressure can be monitored continuously using
extravascular pulse pickups and time-domain techniques to determine the pulse-wave
velocity. Gallaghan et al. studied arterial elasticity with canine common carotid arteries in
1986 and concluded that the Moens-Korteweg equation provides a very good basis for
predicting the increase in PWV with increasing bias pressure.
Being dependent on pressure, Young’s modulus E is not a constant. The relationship is
of the form E = E0eαP, where E0 is the zero pressure modulus, α is a constant that depends
on the vessel, and e is 2.71828 (Geddes 1991, reading 100-103). Thus, Formula (1) can
be expressed as
27
PWV=
tE 0 eαP
.
ρd
(2)
Geddes et al. proposed in 1981 a method where blood pressure pulses will be measured
on the same artery in different locations along the vessel. In this method, the pulses
become detectable, when the cuff pressure applied on the upper arm falls below systolic
blood pressure. Moreover, as the pressure decreases, the pulse delay shortens until it
stabilizes. This point can be used to determine diastolic blood pressure. Geddes used this
method to measure anaesthetized dogs.
In his paper of 1988, Okada showed that PWV correlates well with age, gender,
systolic and diastolic blood pressure and serum phospholipids. He used two sets of
photoplethysmographs to record pulses in the fingertip and toe tip. Later in 1996, Franchi
et al. applied the ECG, a direct blood pressure measurement and photoplehysmography,
PPG, to the ear lobe. Using the first and second derivative of the photoplethysmographic
signal to determine pulse onset, they found a correlation between delays from the R-wave
to the aortic pulse and from the aortic pulse to the ear lobe pulse.
Šantić & Šaban proposed in 1995 a method similar to that presented by Geddes in
1981, but used a finger cuff and a PPG sensor placed on the tip of the same finger. When
cuff pressure drops, the plethysmographic signal can be sensed at the finger tip, and its
delay to the cuff pressure signal will decrease until it stabilizes. This point can be used to
determine diastolic blood pressure. In 1997, Kerola et al. proposed a method which used
an arm cuff with two EMFi pressure sensors under it. The delay change between signals
from these sensors was to be measured by the cross-correlation method. Later on, Nissilä
et al. in 1998 (paper III) and Sorvoja et al. in 2001-2005 (papers IV-VIII) proposed a
method, where an EMFi sensor is used to sense radial artery pulsations and a cuff is
placed around the upper arm to occlude the brachial artery. Diastolic blood pressure
determination was based on either pulse amplitude change or pulse transit time, PTT,
change. In 1999, Šantić & Lacrović put forward a method based on determining diastolic
blood pressure using the weighted values of three methods and systolic pressure using
both the oscillometric and the palpation method, which is based on sensing the first pulse
in a distal sensor.
Since then, pulse wave velocity measurements have attracted a great deal of attention.
During the 1990s and 2000s, it has become a broadly approved method of assessing
arterial stiffness, which has proven to be a good marker for predicting premature death. In
1999, Dr. Roland Asmar wrote a book which summarizes the principles of the method, its
clinical applications and factors influencing pulse wave velocity. In principle, pulse
waves can be measured noninvasively on the skin over the artery between two different
points in the arterial tree. As seen in the formula above, the elastic modulus is dependent
on pressure, allowing it to be determined on the condition that we can measure pulse
wave velocity. The modulus alternates in different positions in the arterial tree depending
on the vascular tonus, i.e. vasoconstriction and vasodilation, which makes it difficult to
measure PWV accurately. In his book, Asmar also mentioned that pulse wave velocity
depends on age, gender, genetic factors, salt intake, heart rate and blood pressure.
28
2.6.1 Photoplethysmography for the pulse transit time method
Since the discovery of the Penaz -Volume clamp method and experience gained of the
Finapres blood pressure measurement method, optical photoplethysmographic, or PPG,
finger pulse detection has been the principal pulse detection method for many
researchers. In 2000, Chen et al. proposed a method based on measuring the beat-to-beat
time interval between the QRS apex in the ECG signal and the onset of
photoplethysmogram in an oximeter sensor placed on a fingertip. The method was tested
with 20 cardiac patients during a surgical operation using an invasive method as a
reference. PTT values were calculated and divided into two bands, high (HF, 0.005-0.04
Hz) and low (LF, DC-0.005 Hz), by digital filtering. Before that, the time scale was
reorganized such that one blood pressure pulse corresponded to one PTT value. The
method used invasive blood pressure recordings for intermitted calibration (systolic DClevel). The estimated values for systolic blood pressure were calculated by summing HFand LF-values, and the obtained accuracy for systolic blood pressure was (0.97 ± 0.02)
mmHg with the error remaining within 10% in 97.8% of the monitoring period. This
result shows that systolic blood pressure changes can be tracked using PTT, if the
calibration pressure can be measured regularly and if the elasticity of the arterial wall is
maintained constant.
In terms of determining PTT values, an important role is played by the site where the
blood pressure pulse is recorded. Sensors can be attached either on the carotid and
femoral, as recommended in PWV measurements, or on the radial artery and fingertip. It
is obvious that highest accuracy can be obtained by using two identical measurement
methods and sensors with identical signal conditioning along the same artery. In 2001,
Maguire et al. proposed the use of PPG for measuring pulsations in the brachial artery
and fingertip. Although the arteries are not the same, they are nonetheless in the same
arterial tree, separated by some artery bifurcations.
Another important factor is determining the timing point of the pulses. Pulse arrival
time measurements (PAT) are widely used not only in the biomedical field, but also in
laser and radar ranging, altimetry and profilometry, optoelectronic distance and
dimension measurements, nuclear spectroscopy and fluorescence life-time spectroscopy,
etc. Usually, one of two methods is used: discrimination of the leading edge and constant
fraction (Myllylä 1976 and Vojnovic 2002). Of these, the first method uses a constant
triggering level, while the second uses a constant fraction of the maximum pulse
amplitude as the triggering level, 33% for instance. As a rule, PWV, PTT and blood
pressure measurements determine timing points from the foot of the systolic rise of the
pulse, i.e., at the end of the diastole. As the foot may not be sufficiently distinct due to
noise, it may be advisable to use line or polynomial fitting and crossing-point detection.
There may be erroneous readings depending on the points taken into account during
fitting.
Two research groups, one led by Prof. Asada at the Massachusetts Institute of
Technology, USA, and the other by Prof. Zhang in Hong Kong, have explored the topic
since the turn of the millennium. Their main results on blood pressure measurements will
be presented in the next paragraphs.
29
Starting with Asada’s research group; Yang et al. and Rhee et al. introduced the ring
sensor in 1998. It comprised a photoplethysmography and a pulse oximetry monitor in a
ring, which could be attached around a patient’s finger for continuous measurements 24
hours a day. In addition, the device contained a microcontroller (PIC16C11) required to
control the pulsing of the LEDs (NIR, 940 nm and red, 660 nm), data acquisition, signal
conditioning and filtering. It also had a MEMS accelerometer for artifact rejection and an
RF transmitter (OOK, on-off-keying) and a button-type lithium battery.
In the next generation of devices (Rhee et al. 2001 and Asada et al. 2003), this single
ring sensor arrangement was replaced by a double, isolating, ring system, where the
inertial mass of the battery and printed circuit board could not influence the inner ring,
where the LEDs and the photo diode were. This system was artifact-resistant and
insensitive for ambient lighting, enabling its application to moving, continuous 24-hour
heart rate monitoring for almost a month using a button cell battery. In addition, with a
discrepancy of only 1.23 pulses/min in terms of rms values, it provided good accuracy.
In 2005, Gibbs & Asada proposed an artifact recovery method, based on a MEMS
acceleration sensor. The method subtracts the signal produced by the MEMS sensor from
the corrupted PPG sensor signal with a certain delay, with the result that the output
exhibits a low distortion level. Although this correction method can be used either in an
additive or logarithmic model, the latter proved better, and it can also be enhanced to
include the Laguerre expansion technique (Wood & Asada 2005).
Shaltis et al. (2005a and 2005b) and Asada et al. (2005) designed a new version of the
photoplethysmography sensor for measuring five vital signs of a patient, including
temperature monitoring. This new device consists of a LED array and a photo diode array
to locate digital arteries. Also, noisy PPG signals can be recovered by a 3-dimensional
MEMS accelerometer and active noise cancellation. All data is transmitted by an IEEE
802.15.4 compliant transceiver. They also considered the possibility of using PPG signals
to determine blood pressure, but found no stationary relationship between PPG and ABP
(arterial blood pressure) over time scales greater than 20 minutes.
Chan et al., members of Zhang’s group, presented in 2001 a method for noninvasive
and cuffless blood pressure measurements for telemedicine purposes. In this method,
ECG was used as a timing base and start signal, and PPG from a finger as a stop signal.
The device included a WAP-based Internet connection. After 20 trial measurements, they
achieved an accuracy of (7.5 ± 8.8) mmHg for systolic, (6.1 ± 5.6) mmHg for mean and
(4.1 ± 5.6) mmHg for diastolic blood pressure using an unspecified commercial cuff
measurement as a reference.
Sometime later, in 2003, Zheng & Zhang published a paper describing measurements
with three different types of PPG ring; convex, flat and concave. A LED and a detector
were fixed on the palm-side of the ring, while the other side housed two ECG electrodes,
one on the inside surface and the other on the dorsal side. A convex-type sensor, in which
the LED and the detector are 1 cm apart on the palm side, proved best in compensating
for motion artifacts. Lee & Zhang also proposed mathematical methods to compensate
for motion artifacts, referred to as stationary wavelet transform and wavelet transform
modulus maxima, achieving a remarkable improvement. Teng & Zhang used this PPG
ring for blood pressure measurements on 15 young, healthy subjects. To provide
reference measurements, they used an oscillometric BP-8800 device manufactured by
Colin, Ltd. Measurements, taken at three different stages: rest, step-climbing exercise and
30
recovery from the exercise, indicated an accuracy of (0.2 ± 7.3) mmHg for systolic and
(0.0 ± 4.4) mmHg for diastolic blood pressure.
A year later, Hung et al. reported a wireless measurement concept using Bluetooth for
telecommunication from a PPG ring to a mobile phone, PC or PDA device. The achieved
accuracy was (1.8 ± 7.6) mmHg for systolic and (0.5 ± 5.3) mmHg for diastolic blood
pressure.
Poon & Zhang extended the measurements to include 85 subjects aged 57 (average) ±
29 (min, max) years, of whom 36 were males and 39 hypertensives. These measurements,
conducted over an average period of 6.4 weeks, resulted in about one thousand pairs of
systolic and diastolic blood pressure estimations being made. Reference results were
provided by the average of results measured by a nurse using the auscultatory method and
results obtained using two clinically approved automated blood pressure meters (BP-8800
Colin, Ltd and HEM-907, Omron, Ltd). Poon & Zhang reached an accuracy of (0.6 ± 9.8)
mmHg for systolic and (0.9 ± 5.6) mmHg for diastolic blood pressure.
2.6.2 Other measurement methods and sensors for
the pulse transit time method
As Geddes says in his book of 1991, “despite the century-and-a-half of research on the
development of a suitable arterial pulse pickup, none exist today that have adequate
fidelity to provide the millisecond resolution needed to employ pulse-wave velocity to
track changes in blood pressure. Perhaps the use of ultrasound to detect movement of the
arterial wall offers the best promise of meeting the requirements for pulse detection”, the
need for high resolution transit-time measurement is essential for accurate measurements.
Since then, technology improvement and miniaturization have given more possibilities
for sensor technology, signal conditioning and powerful real-time data processing.
Multi-channel impedance plethysmography has been used to measure volume changes
caused by blood volume pulsations in a limb (Risacher et al. 1992 and 1997). Initially,
measurements were conducted on a patient’s leg or arm, but later also the finger has
proved a good choice (Šantić and Štritof 1998). One disadvantage of the method is that a
large power supply is needed to feed the electrodes. However, the average current
consumption can be lowered to 0.2 mA by, for example, decreasing the duty cycle to
1:100 (Šantić and Štritof 1998).
Also magnetic resonance imaging has been used for pulse detection (Mohiaddin et al.
1993). In addition, ultrasonic methods have gained wide acceptance in diagnosis, while
papers dealing with pulse wave velocity measurements were first published by the
Japanese researchers Kawabe et al. and Kanai et al. in 1993 and 1994. Subsequently, the
method has been investigated by Stewart et al. and Giller et al. in 1994 and by Hasegawa
et al. in 1997.
Yet another method involves measuring pressure directly using a pressure or force
sensor or a sensor array on the skin, as explained earlier in conjunction with the
discussion on the tonometric method.
Microphones can be used to sense sounds produced by pressure pulses or fluid flows
in arteries. This type of measurement method is referred to as phonocardiography or
31
phonoendoscopy. For example, Jelinek at al (2003) used microphones on the chest and
the left carotid artery.
Optical methods (other than the PPG discussed above) involve fiber optic sensors and
a light source/detector pair either in contact with or close proximity to the skin surface.
Fiber optics can be used to sense pressure (Li et al. 1996) or skin vibrations.
Interferometers have been studied by Hong in 1994 and later by Meigas et al. (2001 and
2003) and Hast et al. (2001-2003, paper IX).
2.6.3 Standards and accuracy validation
In 1987, the American Association for the Advancement of Medical Instrumentation,
AAMI, published a standard for sphygmomanometers, which included a protocol for
evaluating the accuracy of devices (AAMI 1987, revised 1993). In 1990, a protocol was
devised by the British Hypertension Society, BHS, (O’Brien et al. 1990, revised 1993)
and later by the European Society of Hypertension, ESH (O’Brien et al. 2001). This was
then used as the basis for the International Protocol (O’Brien et al. 2002).
According to the AAMI standard, the mean difference between different
measurements must be less than 5 mmHg, or the standard deviation must be less than 8
mmHg with 85% of the measurements and in the 20-250 mmHg range. Accuracy better
than 10 mmHg must be achieved with 95% of the measurements.
In the British protocol, accuracy falls into four grades, A to D, where A denotes the
greatest and D the least agreement. These grades represent the cumulative percentage of
readings falling within 5, 10 and 15 mmHg. To fulfil the BHS protocol, a device must
achieve at least grade B. As the table below indicates, grades in the British protocol are
less strict than in the AAMI standard. Grades are specified separately for diastolic and
systolic blood pressure.
Table 2. Grading criteria used by the British Society of Hypertension.
Grade
Absolute difference between standard and test device
≤5
≤10
≤15
A
60%
85%
95%
B
50%
75%
90%
C
40%
65%
85%
D
Worse than C
Put together by a working group of the European Society of Hypertension, the
International Protocol for Validation of Blood Pressure Measuring Devices in Adults
provides a recommendation for the validation of such devices. Using a standard mercury
sphygmomanometer as a reference, the protocol requires that 33 study subjects, 15 in
Phase 1 and 15 plus 18 in Phase 2. Seven measurements must be performed, starting from
measurements carried out by two observers (reference), which will be averaged, followed
by a supervisor’s test (device under test). Four reference measurement pairs and three
supervisor’s measurements must be conducted on each patient in turn. By subtracting the
mean values of the obtained reference pair measurements from the supervisor’s value, we
32
will arrive at the difference. All subjects should be at least 30 years of age and may be
taking antihypertensive medication. These accuracy criteria are quite similar to those of
the BHS: categories for rounded values are ≤ 5 mmHg (no error of clinical relevance), 610 mmHg (slightly inaccurate), 11–15 mmHg (moderately inaccurate) and ≥ 15 mmHg
(very inaccurate). The table below can be used for validation. For clinical approval, the
device under test must pass all levels.
Table 3. Grading criteria used by the International Protocol.
Phase
Phase 1
Rounded absolute difference between standard and test device
within 5 mmHg
within 10 mmHg
within 15 mmHg
At least one of three
At least one of three
At least one of three
(15x3 = 45 pairs)
Phase 2.1
25
35
40
Two of three
Two of three
Two of three
(33x3 = 99 pairs)
Phase 2.2
65
80
95
All of
All of
All of
60
75
90
2/3 within 5 mmHg
0/3 within 5 mmHg
2.7 Blood pressure values
Blood pressure values are divided into optimal values, normal and high normal values
and hypertension, which in turn falls into stage 1, 2 and 3. For being classified as optimal,
systolic pressure should be less than 120 mmHg and diastolic less than 80 mmHg. This
classification equates with that of the ESH and that of WHO/ISH and is based on clinical
blood pressure values. Other values can be seen in Table 4 (Williams et al. 2004).
Table 4. British Hypertension Society classification of blood pressure (also ESH and
WHO/ISH).
Category
Systolic (mm Hg)
Diastolic (mm Hg)
Optimal
< 120
< 80
Normal
< 130
< 85
High normal
130-139
85-89
Stage 1 hypertension (mild)
140-159
90-99
Stage 2 hypertension (moderate)
160-179
100-109
> 180
> 110
Stage 3 hypertension (severe)
Stage 1 isolated systolic hypertension
140-159
< 90
Stage 2 isolated systolic hypertension
> 160
< 90
When systolic and diastolic blood pressures fall into different categories, the higher
category should be used to classify the patient’s blood pressure level. For example,
160/80 mmHg corresponds to stage 2 hypertension (high blood pressure).
33
2.7.1 Clinically approved blood pressure measurement devices
Table 5 tabulates clinically validated blood pressure measurement devices with their
prices. To meet clinical criteria, the devices must achieve a minimum B grade for both
systolic and diastolic measurements in the revised BHS protocol or pass the accepted
criteria of the International Protocol.
Table 5. Automatic Digital Blood Pressure Devices for clinical use that are also suitable
for home/self assessment. (British Hypertension Society)
Device / Grade
Cost (rrp)
(including VAT)
A+D-767, BHS A/A
£69.99 (June 2005)
A+D-779, International Protocol
£71.36 (June 2005)
A+D-787, International Protocol
£79.99 (June 2005)
A&D UA-774, BHS A/A
£89.99
A&D UA-767, Plus, BHS A/A
£79.99
A&D UA-767PC, BHS A/A
£120.00 (Aug 2005)
A&D UA-767V, With voice to speak reading, BHS A/A
£149.99 (Aug 2005)
A&D UA-767 Plus Memory, With memory capacity to store up to 30 readings, BHS
£84.99 (Aug 2005)
A/A
A&D UA-767P-BT, With Bluetooth output only available to telemedicine service
£176.25 (Aug 2005)
providers, BHS A/A
A&D UA-704, BHS A/A
£39.99 (Oct 2005)
Microlife 3BT0-A, BHS A/A
£59.95 (April 2005)
Microlife 3BTO-A(2), BHS A/A
£79.95 (April 2005)
Microlife 3AG1, BHS A/A
£49.95 (April 2005)
Microlife, As easy as 123, BHS A/A
£44.98 (April 2005)
Microlife 3AC-1 International Validation
£79.95 (April 2005)
Microlife BP A100, Derivative of 3BTO-A
£64.86 (Dec 05)
Microlife BP A100 Plus, Derivative of 3BTO-A
£84.96 (Dec 05)
Omron M5-I, International Protocol
£89.95 (May 2005)
Omron 705-IT, International Protocol
£169.95 (May 2005)
Omron 705-CPII, International Protocol
£149.95 (May 2005)
Omron MX2 Basic (Same algorithm as HEM 737), International Protocol
£52.95 (June 2005)
Omron MX3 Plus International Protocol
£59.95 (June 2005)
Omron M4-I Derivative of 705-IT
Omron M6
£79.95 (Dec 05)
£79.99
3 EMFi pressure sensor array
3.1 EMFi sensor material
Electromechanical film (EMFi), previously known as electro-thermo-mechanical film
(ETMF), has a thin porous polypropylene film structure. This type of film is produced by
injecting gas bubbles into a molten plastic and then extracting a tube of this gas-plastic
mixture with spherical bubbles. The tube is expanded into a thin film by blowing, which
produces biaxially oriented bubbles in it. Next, the film is stretched, which transforms the
bubbles into flat discs with a lateral dimension of 10…100 µm and a vertical dimension
of about 3 µm. The final thickness of the film is 37…70 µm depending on the type of
processing it has undergone. After that, the film is permanently charged either by a plane
electrode corona discharge system in a high electric field or by using electron beam
charging. To provide electrodes, EMFi is metallized on both sides using one of three
methods: sputtering, vacuum evaporation or gluing a metallized thin polyester film, for
instance. In addition to the material itself, the existing patent also covers the
manufacturing method (Kirjavainen 1987, Savolainen & Kirjavainen 1989, Savolainen
1990).
The operation of the transducer is of capacitive nature: when an external force or
pressure introduces mechanical compression into the film, the spatial distribution of
charges within the material changes with respect to the electrode‘s layers. As a result, a
mirror charge proportional to the force is induced at the electrodes. This charge can be
measured with a charge amplifier or a voltage amplifier with very high input impedance.
Biomedical field tests on EMFi include detecting motor activity and breathing
movements in laboratory animals (Räisänen et al. 1992, Heikkinen et al. 1993), but it has
also been used to measure respiration movements and blood pressure pulsations in
humans (Siivola et al. 1993, Kerola et al. 1997). Furthermore, the material has been used
in smart houses to monitor demented elderly people (Lekkala 1996), to recognize walkers
on an EMFi floor (Pirttikangas et al. 2004) and to measure other physiological signals
(Alametsä et al. 2001 & 2004, Junnila et al. 2005, Akhbardeh et al. 2005, Koivistoinen et
al. 2005, Anttonen & Surakka 2005). The film can also be used as an actuator or a
loudspeaker, because it undergoes thickness alterations when a high voltage is applied
35
between the electrodes (Backman & Karjalainen 1990, Backman 1990, Hämäläinen et al.
1996, Antila et al. 1997, Lekkala 1997, Antila et al. 1998, Nykänen et al. 1999, Lekkala
& Paajanen 1999). With a useful spectrum extending as far as 500 kHz, EMFi may also
serve as an ultrasonic transducer (Streicher et al. 2003).
There are two different types of film, standard and highly sensitive, which have a
sensitivity of about 40 and 200 pC/N, respectively. This sensitivity is strongly dependent
on the thermal environment during processing, storage and operation time. Thus, the
charge tends to decrease, if the temperature remains at over 50°C for a long period of
time. Although this characteristic of the material limits its application range, the material
can also be aged to stabilize its sensitivity. The properties of EMFi, including sensitivity
and manufacturing process, have been widely researched and published in recent years
(Paajanen et al. 1999-2002, Peltonen et al. 2000, Neugschwandtner et al. 2001, Raukola
et al. 2002, Lindener et al. 2002, Gerhard-Multhaupt 2002 and Gerhard-Multhaupt et al.
2003, Wegener et al. 2002 & 2003). The results indicate that sensitivity can be enhanced
by means of different gas atmospheres and pressures during corona charging and by using
different gases for bubble filling. EMFi’s temperature range can be extended by
developing the manufacturing processes or changing the base material. As a result, EMFi
is a highly interesting alternative for applications that call for a very sensitive and
inexpensive transducer material that only requires a simple amplifier.
Due to the relatively large gas voids and local corona breakdowns, sensitivity in
different parts of the film may vary. This could lead to problems when an application
necessitates the use of a very small transducer. A case in point is a pressure transducer
measuring blood pressure pulsations over the radial artery in the wrist. In this application,
the transducer must be of the array type, because the correct pulse pressure waveform is
obtained from a single element sitting over the artery. Signals produced by edge elements
may be decayed and noisy due to motion artifacts.
Array transducers based on EMFi have been produced, tested and used to measure the
heartbeat rate of immobile and moving persons (Paper I and Ruha 1998). In addition,
they have been utilized in non-invasive blood pressure measurements on healthy
volunteers and cardiac patients. In these measurements, a cuff was applied to the upper
arm and pulsations were sensed on the distal side of the wrist. Essential findings have
already been published in several conference proceedings and journals (Papers III-VIII,
Vieri-Gashi et al. 2000 & 2001).
In blood pressure measurements of this kind, based on the electronic palpation of the
amplitude and the transit time of pressure pulse waves, pulsations are detected by a wrist
array transducer during the inflation or deflation of the cuff. In the deflation mode, cuff
pressure is rapidly inflated over the presumed systolic pressure level and then slowly
deflated under the diastolic level. The onset of pulsations corresponds to systolic blood
pressure. Diastolic blood pressure, on the other hand, can be determined by the point
where the delay in the pressure pulse sensed by the transducer array stops to diminish
(Geddes et al. 1981, Papers IV-VIII). Though the same method can be used during cuff
inflation or deflation, the latter mode is preferable.
If the transducer array is dense enough to contain four elements per millimetre, it can
be used as a tonometer type of blood pressure transducer array. In this array, the
transducer element that is located over the central part of the artery measures interior
pressure with a high degree of accuracy, dependent mostly on the hold-down pressure and
36
the accuracy of each transducer array element. In addition, this type of measurement also
requires that the transducer’s bandwidth starts at DC. (Geddes 1991)
3.2 Preparation of blood pressure sensor
This study made several modifications to the existing practice of constructing small-sized
transducer arrays. Commonly, transducers are made separately and connected to an
amplifier unit with lengthy transfer lines made of Kapton™ PCB. Such transfer lines,
however, increase input capacitance, which serves to decrease the amplified signal. In the
experiments reported here, the transducer and the transfer lines to the amplifier were both
made of flexible Kapton™ printed circuit board (PCB). The equivalent circuit for EMFi
consists of a charge generator in parallel with a leakage resistor and a capacitor. As
transfer line capacitance and amplifier input capacitance are also parallel, the amplifiers
must be placed as close to the transducers as possible.
A second modification involves exchanging charge amplifiers for CMOS voltage
followers with high input impedance. These amplifiers are simple and do not take up
much room. The transducer array was made by gluing a Kapton™ plate on a plastic base
component using a double-sided adhesive produced by 3M. This plate served as the
ground plane and the second electrode for the transducer. Next, the EMFi material was
glued on top of this plate using the same tape. The following, similarly attached, layer
comprised a ground sealed eight-element transducer electrode plate (Kapton™ PCB). The
dimension of the sensitive area of the transducer arrays was approximately 7x11 mm, as
seen in Fig. 1, which shows the device in its natural size from three different angles.
Transfer lines were soldered to a standard PCB containing the amplifiers. This
configuration has a transfer line capacitance of ~10 pF, paralleling that of the transducer
elements, and its leakage resistance lies in the TΩ range.
Operational
amplifiers
Fig. 1. Eight-element transducer array with EMFi as pressure sensing material. (II)
Such transducer arrays were used to study the sensitivity of sensors both as a function of
temperature and of static and alternating pressure amplitude. The pyroelectrical behaviour
of EMFi manifests itself in offset voltage drift. This offset voltage is affected by the
37
permanent charge of the EMFi material loaded in the transducer, the leak resistance of the
transfer lines and electrodes and the input bias currents of the operational amplifier. Fig. 2
a) presents an ideal operational amplifier with electrostatic discharge (ESD) diodes in its
input, which is common practice for amplifiers with high input impedance. A mismatch
in the properties of these diodes causes an input bias current. If the diodes are exactly
similar, a current is generated only from the positive to the negative power supply pins,
and does not affect the input bias current and offset voltage. A mismatch, on the other
hand, produces a potential that deviates from zero. Fig. 2 b) presents these currents as a
function of potential at two different temperatures. When the temperature rises, the offset
voltage of the unconnected amplifier can be located at the crossing point of the current
curves. Thus, a change in temperature can produce an offset voltage drift.
I2
I1
U
T=T1>T0
T=T0
V1
Uin
V2
Uout
+
Uout
Uin
-
-U
-1.5 V
a)
+1.5V
b)
Fig. 2. ESD diodes (V1 and V2) in the input pins of the operational amplifier, a), and their
influence on bias currents at two different temperatures, b). (II)
When the transducer applies a very low frequency input signal (equal to Uout), the bias
current increases. The bias current to and from the transducer equals the difference
between these curves at the voltage value Uout. In addition, elevated temperature increases
the bias current and may decrease the amplitude of the low frequency signal: the
operational amplifier draws charges to itself.
Another important consideration relates to the sensitivity of the different elements.
Quality variations in the EMFi material and potential gluing faults may produce
sensitivity variations. Such variations are also produced during storage, particularly by
alterations in temperature. Other important factors in this respect include pressure
distribution and cross talk between different channels (array elements). Fig. 3 a) shows
the transducer array, while Fig. 3 b) displays a part of the same array in magnification.
The entire sensor is well shielded to avoid noise. Instead of soldering, the ground
electrode can be glued with an electrical connective glue to avoid introducing a
temperature shock in the EMFi. As the sensor is not in direct skin contact, arterial
pressure is mediated to it by means of plastic heads of exactly the same size as the
transducer element, i.e., 2 mm x 0.6 mm with a 0.4 mm gap between them.
38
Soldering
or glue
Kapton
Lead-through
Kapton foil
electrodes
EMFi film
Operational amplifiers
a)
with
ground
3M double side
adhesive tapes
Printed circuit board with
transducer array electrodes
b)
Fig. 3. Improved version of the EMFi transducer array used to study sensibility and crosstalk between different elements. Fig. b) is a magnification of Fig. a). (II)
3.3 Measurement setups for the blood pressure sensor
Two kinds of pressure generation modules were developed to measure the responses of
the transducer arrays. The first one comprised a computer-controlled electromagnetic
pump capable of producing a pressure pulse with a pre-programmed wave shape. The
system was quite complicated; a PCB (sitting in a personal computer) was used to drive
power circuits in an outside module feeding an electromagnetic membrane pump. This
pump was attached to a metallic chamber with a plastic cover. The cover contained a
cavity, which fitted exactly the transducer array described in Fig. 4. The chamber was
filled with water, and a silicon rubber was used to isolate the water from the transducer
array, see Fig. 4. In addition, the chamber housed a pressure transducer (Sen Sym PS
15GC), whose signal served as feedback-signal to the PCB. After a few pulses, the
system adapted to the desired signal and produced exactly the desired waveform. This
measurement used only a sinusoidal one-Hertz signal.
39
Transducer array with amplifiers
Plastic
cover
Rubber
membrane
Chamber
Pressure chamber
base
filled with water
Fig. 4. Transducer array mounted on the pressure chamber, which produces a sinusoidal one
Hertz signal using a membrane pump. The membrane pump was controlled by a computer
system. Also the temperature inside the chamber was controllable. (II)
As this pressure generation system was incapable of producing high frequency signals,
another pressure generator had to be used for this purpose. The other generator,
consisting of a piezoelectric actuator (Physic Instrument P-840.60), was used to generate
sinusoidal signals for one transducer array element at a time. Contact pressure was
measured from a piezoelectric actuator by means of a force transducer (Honeywell Micro
Switch FSG-15N1A). The size of the contact head equalled that of the transducer
element, 0.6 mm x 2.0 mm, which enabled the investigation of cross talk between the
different transducer elements. In every measurement, a personal computer with a
National Instruments data acquisition board (AT-MIO16 or DaqCard-700) was used for
file readouts. The measurement system is illustrated in Figs. 5 and 6.
40
Oscilloscope
Signal generator
Power source
Force
sensor
amplifier
Piezo actuator
Force sensor
Test head
EMFi test
piece with
amplifiers
Adapter card
Data acquisition board
Fig. 5. Measurement arrangement for measuring the second set of transducer arrays. The
force sensor detects the sinusoidal force generated by the piezo actuator. Both this and the
signal from the EMFi transducer are amplified and then recorded by a personal computer
(PC) using National Instruments data acquisition board. (II)
41
Piezo actuator
Force
sensor
Test head
Translation stage
Aligning
support
PCB mounting
clips
Aluminum base
Fig. 6. Magnified picture of the aligning system presented in Fig. 5 to measure the sensitivity
of the EMFi material. A micro positioning translation stage is used to move the test head just
above the sensor element under test. The test head is aligned exactly on top of the element.
(II)
3.4 Sensor measurements
Firstly, the step response of the transducer was measured using the pressure chamber at
room temperature (22 °C). Fig. 7 presents the step response signal with the time constant
of 1 hour and 10 minutes. In this test, the sensors covered an area of 5 mm2 and the
operational amplifier was TLC2272. Since the time constant yields 38 µHz as the lower
limiting frequency, the sensor material can be considered as suitable for biomedical
applications. It must be noted, however, that the time constant is strongly dependent on
temperature and the amplifier’s input bias current.
Amplitude (percentage
of maximum)
42
100
80
60
40
20
τ
37%
0
0
1000
2000
3000
4000
Time (s)
5000
6000
7000
8000
Fig. 7. Pressure step response of the EMFi transducer using TLC2272 as operational
amplifier at room temperature (21 ºC). (II)
Secondly, the temperature behaviour of the transducer was studied by placing a sensor in
a temperature cabin, while the temperature was decreased from 40°C to 0°C by 5°C
steps. Fig. 8 shows the offset voltages as a function of time. As can be seen, EMFi
exhibits pyroelectrical behaviour. The film produces upward spikes, but the bias currents
drive the voltage down. Consequently, when a sensor based on EMFi is employed to
detect sub-Hertz pressure alterations, temperature should be maintained at a stable level
near room or body temperature.
Offset voltage (mV)
1200
700
40
35
30
25
20
15
10
5
0
200
-300
-800
0
200
400
600
Time (min)
800
1000
1200
Fig. 8. Offset voltages of the eight-element transducer array as a function of time with eight
temperature steps. (II)
This pyroelectric behaviour was further studied by decreasing the temperature linearly
during a one-week period. The thus obtained offset voltages are presented in Fig. 9. As
the figure shows, the offset voltage is almost zero between 25°C and 40°C. This indicates
that the sensor material is suitable for biomedical applications in which the sensor is
either very near or in contact with skin and thus near the body temperature. Nevertheless,
when the sensor is in actual skin contact, problems may arise due to humidity evaporating
43
Offset voltage (mV)
from the skin. Because the transducer is capacitive by nature, its electrodes must be
properly sealed to prevent humidity from affecting the edges of EMFi’s electrode
interface.
1000
0
-1000 0
5
10
15
20
25
30
35
40
45
-2000
-3000
-4000
Temperature (°C)
Fig. 9. Offset voltages of the EMFi transducer array elements as a function of temperature.
(II)
Since the characteristics of the operational amplifier evidently affected the measurements,
the EMFi material was tested by resetting the amplifier’s input electrode to a positive or
negative supply voltage and taking a readout when the voltage was stabilized. Also the
other electrode was connected to the positive or negative supply voltage instead of zero
voltage as usual. The influence of the voltage Ue produced by the other electrode is
clearly indicated by several readings at different temperatures in Fig. 10, whereas the
influence of the reset voltage was very small. In a further test, the transducer was
replaced with a 10 pF polystyrene capacitor, see Fig. 11. As the leakage resistance of the
capacitor roughly equals that of the EMFi material, Fig. 11 presents only the behaviour of
the operational amplifier. The test shows that the EMFi measurement differs considerably
from the capacitor measurement. The characteristics of EMFi seem to dominate over
those of the amplifier at higher temperatures. According to Figs. 10 and 11, the offset
voltage can be compensated for by applying opposite polarity voltage to the “ground”
electrode of the EMFi transducer by means of an integrator, for example.
Offset voltage (mV)
44
1500
1000
500
0
Ue=+1.5V ,reset=+1.5V
Ue=+1.5V ,reset=-1.5V
Ue=-1.5V ,reset=+1.5V
Ue=-1.5V ,reset=-1.5V
-500 10
-1000
-1500
-2000
15
20
25
30
35
40
45
50
Temperature (°C)
Fig. 10. Offset voltages as a function of temperature when different voltages are applied to the
“ground electrode” of the EMFi transducer and output pins are reset with different voltages.
(II)
Offset voltage (mV)
0
-300
10
15
20
25
30
35
-600
-900
40
45
Ue=+1.5V ,reset=+1.5V
Ue=+1.5V ,reset=-1.5V
Ue=-1.5V ,reset=+1.5V
Ue=-1.5V ,reset=-1.5V
50
-1200
-1500
Temperature (°C)
Fig. 11. Offset voltages as a function of temperature when different voltages are applied to the
electrodes of a capacitor and reset. (II)
The sensitivity of the transducer array was measured with a pressure pump chamber. Fig.
12 presents the sensitivities of the eight elements as a function of temperature using a
one-Hz pressure signal with a sinusoidal shape. Although the sensitivity of the various
elements varied considerably, this variation appeared to be independent of temperature.
Next, a set of similar measurements was carried out, but this time sensitivity was
measured as a function of alternating pressure and static pressure. In the alternating
pressure measurements, static pressure was 100 mmHg and alternating pressure varied
from 10...180 mmHg in 10 mmHg steps. In the static measurement, on the other hand,
alternating pressure was 100 mmHg and static pressure ranged from 10...130 mmHg.
Being almost identical to those in Fig. 12, the results will not be presented here. The
average sensitivity of the six transducer arrays that were involved in these measurements
was 6.4 mV/mmHg ±18%. Variation in the sensitivity of the different elements in each
array was smaller (5% to 10%).
Sensitivity (mV/mmHg)
45
10
8
6
4
2
0
25
30
35
Temperature (°C)
40
45
Fig. 12. Sensitivity of the transducer array as a function of temperature. (II)
A second set of transducer arrays was studied by investigating cross talk and quality
(sensitivity deviation) in different parts of a high-sensitivity EMFi sheet. To stabilize its
quality, the sheet was first aged by the manufacturer. Then the sheet, measuring 60 cm x
85 cm, was divided into 4 x 5 equal parts. Little sections were cut out of selected parts in
the horizontal and vertical direction to be used as sensors. The transducer arrays were
made separately for each PCB and measured at the one-Hertz sinusoidal force. Every
measurement was made four times and then averaged. The results of this set, totalling
800 measurements, are presented in Table 6. Due to aging, the sensitivity of this
transducer array set was much lower than that of the previous set, particularly around the
edges of the sheet. Furthermore, the standard deviation of this set was quite high,
indicating that the transducer material is potentially inadequate for certain measurements,
such as measuring the heart rate of a jogger (Paper I, Ruha 1998).
Table 6. Quality study on the EMFi material.(II)
Cut type
Couples
Average sensitivity
Standard deviation
(mV/mmHg)
(%)
Horizontal
13
2.19
±11.3
Vertical
12
2.11
±16.0
Sensitivity was also analyzed as a function of static and dynamic pressure. Compared
with the previous set, this set used considerably higher pressure values. Presented in Figs.
13 and 14, the results point out that, in contrast to the previous set, sensitivity decreased
when static pressure increased. On the other hand, the transducers’ sensitivity to
alternating pressure was stable, which is consistent with the result of the previous set.
Sensitivity (mV/mmHg)
46
3.5
3
2.5
2
1.5
1
0.5
0
0
500
1000
1500
2000
2500
3000
Static pressure (mmHg)
Sensitivity (mV/mmHg)
Fig. 13. Sensitivity of the eight-element transducer array as a function of static pressure at
sinusoidal alternating pressure of 300 mmHg. (II)
3.5
3
2.5
2
1.5
1
0.5
0
0
200
400
600
800
1000
1200
1400
Alternating pressure (mmHg)
Fig. 14. Sensitivity of the eight-element transducer array as a function of alternating pressure
at static pressure of 700 mmHg. (II)
The standard deviation of sensitivity measurements can be improved by laminating two
or more films on top of one another. Hence, five transducer arrays were made by
laminating two EMFi layers. This method produced only a slight improvement in
sensitivity, but had a marked effect on the sensitivity deviation between the elements,
resulting in 2.62 mV/mmHg ± 7.7%.
To measure cross talk, two couples consisting of two types of transducer array were
made by adhering a thick (200µm) or a thin (70µm) aluminium film on an EMFi film,
functioning as a ground electrode. Table 7 and figure 16 show the average values for the
sensitivity, deviation and cross talk of all these transducer types, including also single and
sandwich-type transducer arrays. As can be seen, with 3.34 mV/mmHg ± 5.1%, thin
aluminium produces the best sensitivity and lowest standard deviation and cross talk
values.
47
Table 7. Sensitivity and standard deviation in sensitivity for different sensor types. (II)
EMFi layers and electrode material
Sensitivity
Standard deviation of sensitivity
(mV/mmHg)
(%)
One layer, Copper on Kapton
2.1
13.6
Douple layers, Copper on Kapton
2.6
7.8
One layer, thich aluminium
1.5
7.8
One layer, thin aluminium
3.3
5.1
Cross talk (%)
100
80
60
40
20
0
One layer, Kapton
Douple layers,
Kapton
Thick aluminium
Thin aluminium
Fig. 15. Average cross talk values of all EMFi transducer array types. The black columns are
the signal of the transducer element, where the force is focused and the others are
proportional signals in the neighbouring elements. (II)
4 Blood pressure measurements
4.1 Measurement setups
The measurement device consisted of a standard 13 cm cuff, a wristwatch-type fourchannel pressure transducer array, an invasive catheter (Ohmeda), an amplifier, a filtering
unit and an automatic pressure-controlling unit. A lap-top computer equipped with a data
acquisition board by National Instruments (NI Daq 700) was used to measure the outputs
of the amplifiers. Signals were sampled at the 100 Hz frequency. Based on electrothermo-mechanical film (Kirjavainen 1987, Ficher et al. 1990), the transducer array was
specifically designed for detecting pulsations in the radial artery (paper I & II, Sorvoja
1998, Ruha 1998, Kokko 1999). The amplifier unit was used to amplify and band-pass
filter cuff pressure signals (1-15 Hz, Butterworth filter response, fourth order high pass,
second order low pass) and signals produced by the sensor. These measurements were
made twice, in supine and sitting position, during inflating and deflating cuff pressure
with a one-minute follow-up recording period between them.
The next measurements were based on signals produced by the DATEXTM patient
monitoring system and another, self-made, blood pressure measurement device. In
addition to the device described above, this one also comprised an MSP430TM-based
microcontroller unit for determining blood pressure. Using 64 kBytes, the measurement
and blood pressure determination algorithm was coded in the microcontroller
manufacturer’s assembly language (~1900 rows text). The compiled program used 512
bytes of RAM and 3970 bytes of ROM. Cuff pressure was inductively transmitted to the
microcontroller unit by transmitter-receiver pairs designed by Polar Electro Ltd. Pressure
information was based on time-interval measurements, with one pulse a second
corresponding to zero pressure and two pulses a second to a pressure of 256 mmHg. Also
the sensor’s signal band was slightly different from that of the first measurement system:
four-element transducer signals were amplified and filtered using a first-order band-pass
filter (1.7 Hz-11 Hz) in a full custom integrated circuit (Ruha 1998). A connection from
the DATEXTM device provided an ECG signal and each patient’s radial and pulmonary
intra-arterial blood pressure. An automatic pressure-controlling unit, connected to the
laptop computer, started cuff pressure inflation and sent all pressure data wirelessly to the
49
microprocessor unit. Designed to assess the functionality of the whole system, the
microprocessor unit contained the same blood pressure detection algorithm as the laptop
computer. The data acquisition rate of the recording system was 100 samples/second.
This system is illustrated in Fig. 16, with the connection cable to the patient monitoring
system missing.
Fig. 16. Second measurement system. From left to right: self-controlled pressure pump unit,
amplifier and band-pass filter unit, MSP430-microcontroller unit, cuff for adults with an arm
circumference of 24-33 cm (Speidel & Keller), aneroid pressure metering device with a pump
for pressure calibration and lap-top computer equipped with the National Instrument DAQ
card 700. In the lower right hand corner is the four-element wrist sensor used to detect
pressure pulses in the radial artery.
4.2 Study subjects
The first set of measurements was carried out at the Oulu University Hospital during
1997. Most of the study subjects, 16 in all, were healthy volunteers. In addition, seven
elderly patients were measured. It soon became apparent that the blood pressure signals
of the two groups differed in shape and amplitude. At that time, the blood pressure
determination algorithm focused on detecting changes in pulse amplitude, not in the
recorded delays near the diastolic pressure level. Paper III describes the results of
measurements made on healthy persons, and their characteristics are shown in Table 8
(average values and standard deviations). An invasive catheter (Ohmeda) was used as a
reference blood pressure method. It can be seen that both the men and the women were
50
normal weight, and that their blood pressures were in the normal or high normal range.
The width of the cuff was 13 cm.
Table 8. Study subjects (healthy volunteers). The values are average values ± standard
deviations or range . (VI & VIII)
Item
n
Men
Women
12
4
Age, years
26.2(20 − 33)
30.8 (27 − 34)
Height, cm
180.8 ± 5.6
170.0 ± 2.4
Weight, kg
79.4 ± 10.1
63.5 ± 6.0
BMI, kg/m2
24.2 ± 2.4
22.0 ± 1.9
Arm circumference, cm
31.6 ± 2.6
28.3 ± 2.4
Blood pressure measurements, n
IA systolic BP, mmHg
98
23
129.1±11.2
122.3 ± 10.1
IA mean BP, mmHg
87.3 ± 6.5
83.5 ± 5.7
IA diastolic BP, mmHg
67.4 ±5.0
63.7 ± 5.3
Heart rate, bpm
66.2 ±11.3
69.8 ± 8.1
Taking about 5 months, the next set of measurements was conducted on elderly cardiac
patients at the Oulu University Hospital in 1998-1999. Also in these measurements,
reference readings were provided by the invasive method with the signal obtained in
analog form directly from the patient monitor’s output.
In the end, three to five measurements on 51 patients were saved on the computer for
further analysis. Unfortunately, a number of measurements had to be rejected due to
inadequate recording (missing arm circumference) or insufficient cuff width. As the cuff
width in all measurements was 13 cm, the study subjects’ arm circumference had to be
within 24-33 cm to ensure adequate accuracy (Geddes 1991, manufacturer’s claim). Table
9, tabulating the characteristics of these eligible patients, shows that their arm
circumference was appropriate and that the invasive blood pressure values were within
the optimal normal range.
It has to be pointed out that the International Protocol had not been implemented at the
time of the study, and could therefore not be followed. In addition, we did not use a
noninvasive method as a reference, but an invasive one, but this is not a problem, as
invasive methods invariably yield more reliable results. Also the number of study
subjects (16 healthy and 20 patient) exceeded the requirements of the International
Protocol (33 in all) making the study statistically valid. Of course, both sexes should have
been represented in equal numbers.
51
Table 9. Test subjects (cardiac patients). To provide a single value for each patient, the
intra-arterial (IA) blood pressure (BP) values given are average values of the measured
IA diastolic, mean and systolic BP values (average values ± standard deviations or
range). (VI & VIII)
Item
Number
Men
Women
14
6
Age, years
67.8 (59−75)
65.2 (49−75)
Height, cm
168.0 ± 8.1
163.0 ± 6.6
Weight, kg
71.4 ± 11.8
74.1± 10.0
Body mass index, kg/m2
25.4 ± 4.4
28.0 ± 4.6
Arm circumference, cm
32.3 ± 1.0
32.7 ± 0.5
Bypass operation
13
4
Valve operation
3
2
Blood pressure measurements, n
40
13
114.8 ± 21.3
118.3 ± 19.9
IA systolic BP, mmHg
IA mean BP, mmHg
77.8 ± 9.2
79.6 ± 14.8
IA diastolic BP, mmHg
59.6 ± 6.1
59.6 ± 14.5
Heart rate, pulses/min
87.2 ± 13.5
81.5 ± 5.4
4.3 Blood pressure determination in the electronic palpation method
As shown in Fig. 17, systolic blood pressure can be easily determined at the point where
the last (or first with deflating cuff pressure mode) radial artery pulse appears in the wrist
transducer. We used an array type of transducer to facilitate the placement of the wrist
transducer, as the array element with the strongest pulsation is easy to pick out. The
figure presents the IA blood pressure signal, band-pass filtered signal and electronically
palpated (EMFi sensor) signal at increasing cuff pressure levels. The filtered IA blood
pressure signal has exactly the same bandwidth as the electronically palpated signal.
Surprisingly, the palpated signal does not begin to decrease after the point at which
cuff pressure exceeds diastolic blood pressure. In fact, the signal amplitude slightly
increases, which is why the point where the amplitude signal begins to decrease cannot
be used to determine diastolic blood pressure. Consequently, diastolic pressure is
determined at the point where the pulse delay begins to increase, and systolic blood
pressure at the point where the last pulse is detected. The cuff pressure at these time
points corresponds to diastolic and systolic pressure.
High-pass filtered Intra-arterial and
intra-arterial
cuff pressure (mmHg)
pressure (mmHg)
52
150
100
50
0
5
10
15
Time (s)
20
25
5
10
15
Time (s)
20
25
5
10
15
Time (s)
20
25
20
0
-20
High-pass filtered
signal of EMFi
pressure sensor
-40
20
0
-20
Fig. 17. Determination of systolic and diastolic blood pressure in the electronic palpation
method.
Diastolic blood pressure determination is based on pulse transit time change. When
pressure in the cuff exceeds the diastolic pressure level, pulses on the distal side of the
cuff will be delayed. To accurately measure the point where this delay begins requires an
accurate timing base. The best timing reference is afforded by the radial artery of the
other hand, and, if appropriate data processing is performed, the time difference can be
easily determined.
Standard methods, such as the leading edge or constant fraction method, can be
applied to the determination of diastolic blood pressure. Preliminary calculations with
MatlabTM showed that the leading edge method is not accurate due to variations in pulse
amplitude and profile. Therefore, the constant fraction method was applied to the cardiac
patient group in the 20% to 60% range using 5% steps as the trigger level. It was shown
that, when the triggering level was between 30% and 45%, the average diastolic blood
pressure error remained stable at the 2.5 mmHg level. Also the standard deviation of the
errors was calculated for the whole patient group. The average of standard deviation as a
function of triggering level was shaped like parable opening upwards and reaching its
lowest point at the 35% level. These findings are presented in Figs. 18 and 19, using three
different determination methods. The peak of the first derivative, varying from 22% to
50%, had an average level of 37%, which implies that to minimize error, timing should
be based on the level with the fastest rising or sloping signal. This leads to the conclusion
53
that it is advisable to use derivated signals of blood pressure pulses in the right and left
hand. In this situation, signals can be measured invasively using a catheter and
noninvasively using a pressure sensor array.
Diastolic error (mmHg)
4,0
3,0
2,0
1,0
0,0
-1,0
20
25
30
35
37
45
50
60
method 1
method 2
method 3
-2,0
-3,0
Triggering level (%)
Standard deviation of
diastolic error (mmHg)
Fig. 18. Averages of diastolic blood pressure error with the cardiac patient group as
determined by three different methods.
12.0
10.0
8.0
method 1
6.0
method 2
4.0
method 3
2.0
0.0
20
25
30
35
37
45
Triggering level (%)
50
60
Fig. 19. Standard deviations of diastolic blood pressure error with the cardiac patient group
as determined by three different methods.
Figure 20 presents a flowchart with typical wave shapes for calculating and determining
the time delay between pressure pulses from both hands. Firstly, IA blood pressure and
EP signals are filtered, and the transducer element with the strongest amplitude in the EP
transducer array is chosen. After that, the signals are interpolated by a factor of ten to
obtain a better timing reference. Both signals are then derivated and the time difference is
determined for each pulse on the basis of correlation calculations. The gained signals are
then filtered using three-point moving median and average filtering. Next, the signals are
interpolated so that every cuff pressure integer value has a corresponding pulse delay
value. Specific “triangle conversion”, described later, is performed at the following stage
to find the cut-off point for line fitting, which is used to establish the point, at which the
pulse delay begins to increase. This point indicates the diastolic blood pressure value.
54
IABP
(catheter)
EP
sensor array
Cuff pressure
Software BP
filtering
(1-14 Hz)
Hardware BP
filtering
(1-14 Hz)
Moving
average
filtering
Extra filtering
Extra filtering
Interpolation
dp/dt
dp/dt
Interpolation
Interpolation
1
Time difference determination based
on cross-correlation calculations of
the signals
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
0
200
400
600
800
1000
1200
1400
1600
3 points moving median filtering
3 points moving average filtering
Interpolation
“Triangle conversion “
Line fitting
Crossing point detection
Diastolic blood pressure determination
Systolic blood pressure determination
Fig. 20. Flowchart diagram for the electronic palpation method. (VIII)
Since the EP blood pressure signal in the radial artery has undergone analogue
hardware filtering, the IA pressure signal must be filtered similarly. Further, to eliminate
55
noise that is coupled to the sensor signal from the radial artery, we used additional
software band-pass filtering. Owing to the 100 Hz sampling rate used in data acquisition,
the data signals were interpolated for better timing accuracy. According to our
experience, an adequate resolution level is achieved at a ten-fold sampling rate. After
interpolation, a derivated signal is calculated, and the time difference of these signals is
worked out on the basis of correlation calculations, i.e., the best correlation corresponds
to a situation where the signals overlap one another exactly. Having clearly and
accurately determined the time difference, the next step involves locating the point where
the pulse delay begins to increase. This turning point can be accurately determined by
means of “triangle conversion” after median and moving average filtering and
interpolation. In this conversion method, all time values are recalculated to form a
triangle, and line fittings are used to establish the accurate turning point. To this end, the
time values have to be inverted to obtain decreasing values, and Equation (3) can be
applied to attain new values
Tnew (i)=Told (i)-
( Told (1)-Told (n) ) (i-1)
n-1
-Told (n) .
(3)
In this equation, Tnew(i) is the new time value for the index (i), Told(i) the old value and (n)
the index value for the last measured time value. In Fig. 21, Fig 21.1 illustrates the
increase in pulse delay during the measurement shown in Fig. 17. The pulse time
difference values is then inverted (Fig 21.2) and filtered using three-point moving median
and average filtering (Figs 21.3 and 21.4). In Fig 21.5, the data is interpolated and in Fig
21.6, “triangle conversion” is performed to find the diastolic blood pressure. The painted
dark areas corresponds each others. Line fitting was used to locate the point at which the
line crosses the 100% level. This point correlates to diastolic blood pressure; see the
arrow in Fig. 21.5. This algorithm provided the value of 67 mmHg for diastolic blood
pressure, while the intra-arterial value alternated between 67 and 72 mmHg during the
measurement. The last detected pulse corresponds to systolic pressure and the obtained
value, 157 mmHg, is within the systolic pressure range, 152 to 162 mmHg. As seen, the
line fitting ranges are clearly defined by the turning point (highest value) and, after that,
the crossing point value of the fitted bolded line and the 100% level line can be
calculated. This point corresponds to diastolic blood pressure in the low-pass filtered cuff
pressure.
56
100
5 10
0
3.
-100
0
50
5 10
ms
ms
-50
0
100
0
5.
-50
0
50
0
ms
1.
15 20
s
25 30
-100
0
50
0
ms
0
15 20
s
25 30
100
mmHg
150
2.
5 10
15 20
s
25 30
15 20
s
25 30
4.
-50
0
100
5 10
50
6.
%
ms
50
6.
0
0 20 40 60 80 100 120 140
mmHg
Fig. 21. Pipeline determination of diastolic blood pressure from the time difference between
the filtered and derivated signals of intra-arterial and electronically palpated blood pressure
pulses. 1) Time difference, 2) inverted time difference, 3) 3-point moving median filtered time
difference, 4) 3-point moving average filtered time difference, 5) interpolated time difference
and 6) “triangle conversed” time difference. (VIII)
Pressure pulse wave shape, for example augmentation, can be used to assess the level of
arterial stiffening. The problem is that augmentation depends on the systolic pressure
level and appears and disappears with pressure fluctuations. Fig. 22 presents a typical
waveform for an augmented pressure pulse. Systolic pressure is boosted some 20% over
the original value, due to reflections from the periphery. The manifestation of this
augmentation is easily tested by measuring the rising time from 10% to 90% of the pulse
pressure (Trise), or the time from the 50% level to the top level (T2) or the pulse above
the 80% level (T1).
The original systolic pressure can be defined by the “systolic turning point” of the first
derivative signal after the peak (Fig. 23). Unfortunately, the signals are often weakened
by noise. The second derivative can also be computed, but subsequent signals are too
noisy. It was established that the percentage of the first derivative peak of the pulse
pressure is well suited to describing arterial stiffness. Hence, if the pressure slope is
highest shortly after the foot of the pulse (diastolic value), the patient is likely to suffer
from stiff arteries. In young and healthy persons, however, this level is near the 50% level
57
Intra-arterial pressure (mmHg)
of the pulse pressure. Another way of describing the waveform is the percentage of the
pressure pulses’ mean pressure level, which can be easily computed.
Also pressure pulse transit time was defined, followed by a calculation of pulse wave
velocity (PWV). Normally, PWV is noninvasively measured using pressure pulses
between the carotid and femoral artery, followed by a determination of the distance
between the measuring points (Asmar et al. 1995 and Asmar 1999). In this study, time
was determined from the ECG’s R-spike to the end of diastole, when the pressure is just
about to rise again. Distance from the heart to the wrist was defined afterwards by
multiplying the patients' height by the coefficient of 0.43, an empirically established
mean value for normal young people (according to Dr. Jukka Hast). This value shows an
individual variation of about ±10%, producing a corresponding error in the determination
of PWV.
150
100
100%
90%
80%
T1
50%
T2
10%
Trise
0%
50
0
2500
3000
Time (ms)
3500
4000
Fig. 22. Blood pressure signal and specific time values producing a numerical signal
description. (V)
58
200
Pressure (mmHg, a.u.)
Systolic pressure level
150
100
Mean pressure level
First deriv. peak level
Diastolic pressure level
50
0
-50
2500
3000
Time (ms)
3500
4000
Fig. 23. Blood pressure signal (upper) and its first derivative multiplied by five (below). (V)
Other factors were also tabulated, including the visually detected existence of
augmentation, average of intra-arterial systolic mean and diastolic pressure, age, weight,
height, cardiac output and peripheral and blood temperature.
4.4 Accuracy of the electronic palpation method
Systolic pressure was determined visually from the last palpated pulse, while diastolic
blood pressure readings were obtained by the time delay change method described above.
Reference values were provided by the average IA systolic and diastolic pressure. The
corresponding accuracies are presented in Table 10. For the healthy group, the measured
accuracies are inadequate for fulfilling the AAMI standard. On the other hand, the
deflating mode yielded better results for diastolic pressure and the inflating mode for
systolic pressure. For the cardiac patient group, the mean error was (-0.6 ± 5.1) mmHg
for diastolic and (0.7 ± 6.3) mmHg for systolic pressure, which are considerably better
than the accuracies obtained with the healthy group and fulfil AAMI standard. This can
be explained by the pulse rising time: the larger the rise time, the larger the pulse delay
change, which allows a more accurate determination of diastolic blood pressure.
59
Table 10. Accuracy (average values ± standard deviations) of the diastolic and systolic
blood pressure measurements in the healthy and the cardiac patient group using the EP
method. The reference value is the average value of IA diastolic, mean and systolic blood
pressure. (VIII)
Mode
Mean errors
Diastolic
Systolic
(mmHg)
(mmHg)
Supine
−5.5 ± 5.2
−1.7 ± 7.9
Sitting
−11.2 ± 6.1
−3.2 ± 6.7
Supine
2.0 ± 6.6
−6.3 ± 9.2
Sitting
0.1 ± 7.3
−7.1 ± 8.3
Supine
−0.6 ± 5.1
0.7 ± 6.3
Healthy
Inflating
Deflating
Patients
Inflating
Other possible determinants for systolic, mean and diastolic blood pressure errors were
examined by means of correlation calculations. Correlations between pressure errors and
IA pressures were fairly low, the only exception being the correlation between IA
diastolic pressure and EP diastolic pressure error, which was 0.54. We also noted that the
pressure errors were unaffected by the transducer array’s amplitude level.
Table 11 presents the above-mentioned indices and the corresponding correlations
between the obtained systolic blood pressure errors. The indices are shown in a
descending order according to the correlation coefficient. It can be seen that features of
the rising edge of the pressure pulse correlate better with the systolic error than those of
the falling part. The first derivative peak level of each pulse pressure describes the pulse's
highest rise rate.
The table also indicates that many features have no correlation with systolic pressure
error. If the absolute value of the correlation coefficient is less than 0.3, the correlation is
considered negligible. Thus, there seems to be an insignificant correlation between
systolic pressure error and factors like pulse wave velocity, oxygen saturation, cardiac
output, body mass index, peripheral temperature, diastolic and mean pressure, age and
weight.
60
Table 11. Correlations between systolic blood pressure error and different indices of the
pressure pulse waveform. All values are average values. (V)
Correlation coefficient R
First derivative peak level of pulse pressure
-0.52
Rising time (10%…90%)
0.49
Time from 50% of pulse pressure to the maximum
0.46
(MAP-DIAS)/(SYS-DIAS)x100%
0.44
Time from diastolic to systolic pressure (0%…100%)
0.40
Systolic pressure (last palpated pulse)
0.35
Arm circumference
0.32
Augmentation (visual)
0.32
SYS, Systolic pressure (intra-arterial)
-0.31
Pulse width at 80% level of pulse pressure
0.27
Electronically palpated amplitude
0.21
Weight
0.17
Age
0.16
MAP, Mean arterial pressure
-0.14
DIAS, Diastolic pressure (intra-arterial)
-0.13
Peripheral temperature
-0.12
BMI, Body mass index
0.11
Height
0.07
PCWP, Pulmonary capillary wedge pressure
-0.06
CO, Cardiac output
0.05
SpO2, Saturation of oxygen
0.04
PWV, Pulse wave velocity
0.03
Blood temperature
0.00
Linear regression can be used to diminish the dependence of each index. The index
having the highest correlation index was the first derivative peak level of the pulse
pressure. Without correction, the linear regression equation for the electronic palpation
method vs. intra-arterial method is y = 0.79x + 29.05 and the correlation coefficient R2 =
0.61. Fig. 24 displays the first derivative peak level of the pulse pressure method, a
corrected scatter gram, the equation and the coefficient, which all show improvement.
The palpated signal can be corrected likewise: if the measuring bandwidth is, say 0.2-15
Hz, the palpated signal is not distorted and the waveform closely resembles that of the
intra-arterial pressure pulse. Hence, this accuracy enhancement method can be easily
adapted to the noninvasive electronic palpation method.
61
Corrected el. palpated
systolic blood pressure
(mmHg)
200
y = 0.84x + 19.17
R2 = 0.68
150
100
50
0
0
50
100
150
200
Intra-arterial systolic blood pressure (mmHg)
Fig. 24. Corrected systolic blood pressure values obtained by the electronic palpation method
as a function of intra-arterial systolic blood pressure. (V)
4.5 Blood pressure determination in the oscillometric method
In this method, the cuff pressure signal p(i) is first filtered using two-second moving
average filtering, i.e., every sample is averaged 99 samples back and 100 samples
forward using the formula:
p(i) = mean[p (i-99 : i + 100)],
(4)
where p(i) is the value of the sample and i the corresponding index. The resulting
pressure signal ignores one hundred samples from the beginning and another hundred
from the end, which has to be considered in later calculations. The filtered signal is then
subtracted from the original pressure signal yielding a high-pass filtered oscillation signal
with negligible attenuation and phase distortion. This procedure allows the study of the
potential effects of analogue filtering on the oscillation curve. The moving average
filtered signal serves as a cuff pressure reference signal. To assess the influence of
filtering on characteristic ratios, Fig. 25 presents both the analogically filtered (HW
filtering) and the amplified oscillation signal as well as the signal gained by subtracting
the two-second moving average filtered signal (SW filtering) from the original cuff
signal. Cuff pressure was first low-pass filtered (0-14 Hz, fourth order) and then HW
band-pass filtered (1-15 Hz for the healthy group, 1.7-11 Hz for the cardiac patient
group) and, finally, amplified to an appropriate signal level. The lowest SW-filtered
signal was gained by subtracting the two-second moving average low-pass filtered cuff
oscillation signal from the original cuff pressure signal. The thus obtained signal is a
high-pass filtered signal with a very low cut-off high-pass frequency. The signal is then
amplified to an appropriate signal level using a software multiplier. Oscillation
amplitudes are determined by subtracting diastolic pressure from systolic pressure (the
62
lower OSC signal). Alternatively, the following lowest value can be subtracted from the
highest value (the upper OSC signal) to get an amplitude value with pulsation amplitude
levels and the corresponding cuff pressures.
150
Cuff oscillations (a.u.)
IA blood and cuff pressures (mmHg)
Intra-arterial blood pressure Cuff pressure
100
50
HW filtered oscillation
0
SW filtered oscillation
0
5
Amplitude determination
10
Time (s)
15
20
25
Fig. 25. From top to bottom: IA blood pressure, increasing cuff pressure, analogically
hardware-filtered (HW) and software-filtered (SW) cuff oscillation signal (with bias) as a
function of time. (VIII)
The oscillation amplitude values can be reconstituted by multiplying the amplitude values
by a multiplication table in accordance with the formula
k(j)=
mean[p p (first:last)]
p p (j)
,
(5)
where pp(j) is IA blood pulse pressure corresponding to the oscillation pulse number j.
This method allows the reconstruction of the oscillation amplitude signal so that it is not
affected by variations in IA pulse pressure; in other words, the pulse pressure amplitude
remains constant.
The oscillation amplitude signal must then be filtered to eliminate the effects of
motion artifacts. Our experiments showed that three-point moving median filtering is
adequate for the purpose. Because the cuff pressure curve is not linear in neither the
inflating nor the deflating pressure mode, the oscillation signal must be normalized. After
normalization, the amplitude signal is multiplied by a constant to attain the maximum
value of 100. Also the corresponding pressure signal is multiplied by a scaling multiplier
such that the highest point of the oscillation curve, or 95% of the maximum, corresponds
to the average mean IA pressure. Also Geddes et al. (1983) used a similar scaling factor.
63
As mentioned earlier, this “highest point” should be on the diastolic side of the oscillation
curve (Ursino and Cristalli, 1994). Fig. 26 presents the normalized oscillation curves
before and after multiplication and the average diastolic, mean and systolic blood
pressure value as well as the value of the scaling multiplier. The normalized amplitude
signal is then interpolated such that a single integer pressure value corresponds to the
oscillation amplitude value. If straight lines are drawn up from the averaged diastolic and
systolic IA pressure values, they cross the normalized interpolated curve at the
characteristic diastolic and systolic points (the CR values in Fig. 26). The measured
patient was a 66-year-old woman, who was 172 cm tall, weighed 75.2 kg and had an arm
circumference of 33 cm. The width of her pressure cuff was 13 cm. First, the average
values for the mean intra-arterial systolic and diastolic blood pressure are calculated.
Then, the scaling multiplier is obtained by multiplying the cuff pressure signal by the
oscillation amplitude, to attain the 95% amplitude level of the average mean blood
pressure. These characteristic ratios are the levels at which the multiplied signal obtains
the averaged systolic and diastolic blood pressures values.
100
80
Scaling multiplier=0.85
60
Systolic CR=49
40
0
0
20
40
60
80
Cuff pressure (mmHg)
IAsys=93
20
IAmean=65
Diastolic CR=32
IAdias=51
Normalized oscillation (%)
95% level
100
120
Fig. 26. Determination of the scaling multiplier and characteristic ratios (CR values) for
diastolic, mean and systolic blood pressure from the normalized cuff oscillation curve in the
oscillometric method. (VIII)
Fig. 27 shows a flowchart of all mathematical operations performed on the data to get the
systolic, mean and diastolic blood pressure values. First, cuff pressure is sensed by a DCcoupled transducer. The signal is then amplified and band-pass filtered using both HW
and SW filtering. Amplitude values are subsequently obtained through specific amplitude
determination for both the HW and SW signals. The data values are filtered using threepoint moving median filtering to eliminate casual motion artefacts. Next, by multiplying
every single pressure value by a multiplier acquired from the IA blood pressure
measurement, the data values are normalized such that 100% marks the highest value. In
addition, the data values are filled to match every pressure integer value with a
64
corresponding normalized amplitude value (interpolation). Now, both signals are
multiplied by the previously determined scaling multipliers so that the 95% level
corresponds to the mean intra-arterial blood pressure value. Finally, the mean diastolic
and systolic blood pressure values are determined separately for both the HW and SW
filtered signals using the previously determined characteristic ratios.
Cuff pressure sensor amplifier
Hardware BP filter (1–14 Hz)
Moving average LP filter
Original signal–filtered signal
Peak
detection
Peak
detection
Amplitude
calculation
Amplitude
calculation
3 points moving
median filtering
3 points moving
median filtering
Amplitude
normalization
Amplitude
normalization
Pressure
interpolation
Pressure
interpolation
Pressure
normalization
Pressure
normalization
Mean blood pressure determination
Diastolic blood pressure determination
Systolic blood pressure determination
Fig. 27. Flowchart diagram depicting the oscillometric method. (VIII)
65
4.6 Accuracy of the oscillometric method
The oscillation method described earlier was used to determine characteristic ratios
separately for the healthy volunteer and the cardiac patient group. For the first group, also
the measurement mode (i.e., inflating and deflating cuff pressure) and position (either
supine or sitting) were considered. Table 12 tabulates the average values for these factors
along with the standard deviations and scaling multipliers for both groups. As the table
shows, the scaling multiplier for the healthy group is about one in every group. In
addition, there is only a slight deviation between measurement mode and position. The
characteristic ratios were found to range from 40% to 48% for diastolic blood pressure
and from 49% to 59% for systolic blood pressure in the different modes. It should be
noted that the values for the diastolic characteristic ratios differ considerably from the
mean values reported by Geddes (82%), Drzewiecki (72%) and Ursino (70%). Geddes
conducted his measurements on humans, while Drzewiecki and Ursino also used
mathematical models and a simple blood pressure signal mode. One explanation to
account for the observed discrepancy between their results and the ones reported here is
the different oscillation filtering bandwidth used in our experiments. The characteristic
ratios for systolic blood pressure obtained in the present study lay within the same range
as those reported in the earlier experiments (55%, 59%, 52%, respectively). Another
interesting finding is that the standard deviation of the characteristic ratio for diastolic
blood pressure was only about 50% of that calculated for systolic pressure.
With the cardiac patient group, the scaling multipliers were less than one, 0.86 or 0.87,
depending on the filtering mode. The average values of the characteristic ratios were
somewhat lower for systolic blood pressure and the standard deviations slightly higher
than the corresponding figures for the healthy group. The oscillation amplitude curve was
gained by subtracting the two-second moving average filtered cuff oscillation from the
unfiltered oscillation curve (SW filtering), which resulted in a low degree of distortion.
As analogically filtered signals (HW filtering) gave almost identical characteristic ratios,
filtering did not have a great effect on accuracy.
Table 12. Characteristic ratios (mean values ± standard deviations) for diastolic and
systolic blood pressure in the healthy and cardiac patient group. (VIII)
Mode
Characteristic ratios, %
Scaling multiplier
Diastolic
Systolic
Supine
39.6 ± 7.4
48.5 ± 11.6
0.96 ± 0.05
Sitting
45.7 ± 9.4
55.2 ± 16.7
1.01 ± 0.08
Supine
44.3 ± 6.6
51.2 ± 13.2
1.02 ± 0.05
Sitting
48.1 ± 8.7
58.5 ± 15.3
1.08 ± 0.06
Supine
45.8 ± 15.6
40.6 ± 12.3
0.87 ± 0.07
Supine
43.9 ± 10.4
47.1 ± 13.9
0.86 ± 0.07
Healthy
Inflating
Deflating
Patients
Inflating,
HW filtering
Inflating,
SW filtering
66
These average characteristic ratio values were used to determine both systolic and
diastolic pressure as well as pressure errors for both groups. The reference values were
averaged from the IA pressure waveforms during cuff inflation or deflation, and the mean
arterial pressure was determined at the point where the cuff oscillation reached 95% of its
maximum amplitude curve. The results are presented in Table 13. The pressure error in
the healthy group was found to be fairly low, except in the case of sitting test subjects
measured during deflation of the cuff, which is the standard procedure used in clinics.
This may be due to hydraulic head, i.e., a difference in blood level between the reference
transducer (catheter) and the cuff. For the patient group, however, the average errors were
high, producing overestimates in excess of 10 mmHg, which is outside the standard
range. It must be pointed out that the width of the cuff used in these measurements was
40% or more of the subjects’ arm circumference, thus fulfilling the requirements of the
standards. In both groups, the standard deviation of errors for systolic blood pressure was
too high to fit the standards, regardless of cuff pressure mode and posture. On the other
hand, in measurements of diastolic and mean artery pressure, these deviations were less
than 8 mmHg, which fulfils the requirements.
Table 13. Accuracy (mean values ± SD) of the diastolic, mean and systolic blood pressure
measurements in the healthy and cardiac patient group using the OSC method. The
reference value is the average value of IA diastolic, mean and systolic blood pressure.
(VIII)
Mode
Posture
Mean errors
Diastolic
Mean
Systolic
(mmHg)
(mmHg)
(mmHg)
Healthy
Inflating
Deflating
Supine
2.8 ± 4.5
4.1 ± 5.0
4.0 ± 8.4
Sitting
−0.3 ± 5.4
0.1 ± 6.9
−1.9 ± 10.1
Supine
−1.1 ± 5.1
−1.4 ± 3.7
−3.3 ± 8.5
Sitting
−4.6 ± 6.3
−6.5 ± 4.5
−9.0 ± 11.0
Supine
12.7 ± 6.1
13.4 ± 6.8
15.0 ± 11.6
Supine
10.0 ± 5.3
12.3 ± 7.0
10.2 ± 12.0
Patients
Inflating,
HW filtering
Inflating,
SW filtering
5 Discussion
The historical review showed that, for centuries, blood pressure has been of great
importance both for human health diagnostics and device development. Yet, almost all
clinically approved methods are based on measuring counterpressure and involve the use
of a cuff. This is equally true of “the gold standard”, i.e., the auscultatory method, and the
newer oscillometric method, as well as of the volume clamp method. Some methods,
such as the tonometric and the PTT method (based on pressure pulse transit time), operate
on another principle than counterpressure. Nonetheless, they still require a cuff. In
addition, to work accurately, tonometers demand an accurate hold-down pressure, which
should be controlled periodically. As to the pulse transit-time method, it is also calibrated
using a cuff, but, like tonometers, it can measure blood pressure continuously, pulse-bypulse. This characteristic makes it an attractive method for diurnal, 24-hour, recordings. If
artefact cancellation is added to the measurement system, even accurate ambulatory
recordings can be achieved.
According to O’Brien’s review paper, published in 2001, only few commercially
available blood pressure measurement devices have passed both the AAMI and the BHS
standard, and can thus be recommended. Since then, some of them have passed at least
one of the common standards, AAMI, BHS or International Protocol, of which the oldest
one, AAMI, requires an accuracy of (5 ± 8) mmHg. None of the recommended
noninvasive methods is suitable for continuous measurement. According to Guelen’s
results from 2003, FinometerTM fulfilled the requirements after two sequential
corrections. However, the conclusion was not supported by Elvan-Taspinar et al. in the
same year. This method may be not suitable for long-term measurement, because it can
be painful in the long run, and it is certainly unsuitable for ambulatory measurements,
owing to high power need and artefact sensitivity. Jagomägi et al.’s modified
oscillometric measurement system has the capacity to make continuous measurements
with an accuracy equalling that of Finapres, Finometer and Portapres. All these methods
are also very sensitive for vasoconstriction in peripheral vessels.
One of the most interesting techniques is the pulse transit-time method, which can be
used for continuous noninvasive blood pressure measurement. In this method, a cuff is
used for calibration and variations in blood pressure are recognized on the basis of rapid
changes in pulse transit time. In addition, the application of an adequate motion artifact
68
cancellation method makes it suitable for long-term measurements. Artifact cancellation
can be achieved, for example, by using new, low-power multi-axial acceleration MEMS
sensors. Recent results by Poon & Zhang in 2005, who used the PPG method to detect
blood pressure pulsations in a finger and achieved an accuracy of (0.6 ± 9.8) mmHg and
(0.9 ± 5.6) mmHg for systolic and diastolic pressure, respectively, demonstrate that the
method has great potential for wearable blood pressure monitoring. We may conclude
that since a large majority of recent papers deal with it, the PPG method is the most
popular among research groups. Unfortunately, this method requires power to transmit IR
or visible light into finger tissue. This requirement can, however, be reduced by
decreasing the duty cycle of the transmitted light. In addition, to avoid background noise,
the measurement demands high frequency pulses, which usually results in a need to apply
either high-order band-pass filtering or synchronous detection. Thus, circuit complexity
can be considered to increase, with a resultant increase in power consumption.
EMFi material proved applicable to detecting blood pressure pulsations on the radial
artery and can also be used for pulse shape and velocity measurements. As shown in
Paper VI, the material has the ability to distinctly measure pressure pulsations on
subcutaneous arteries; the average correlation coefficient between intra-arterial and
noninvasive radial pressure pulsations was 0.98, using identical filter bandwidths.
EMFi sensor material needs an operational amplifier operating as a transducer for
detection. Power consumption is not a problem, however, since it is possible to use
EMFi material in conjunction with an ultra low power operational amplifier, such as the
TLC1079 by Texas Instruments, which typically consumes only 10 µW per channel.
Disadvantages of EMFi material are sensitivity variations and decreased sensitivity as
a function of time. In addition, a total disaster is possible, if a certain temperature, the socalled Curie temperature, is exceeded. EMFi material can be used successfully as a
transducer at room temperature, making it a feasible alternative for hospital, point of care
and homecare environments. Also, if a modern low-power acceleration MEMS sensor is
applied on the sensor array package, dynamic hydraulic pressure noise caused by hand
movements and other motion artefacts can be cancelled or considerably reduced.
By fulfilling the AAMI standard, the electronic palpation method proved accurate with
the elderly cardiac patient group. With the healthy volunteers, however, the method only
passed the requirements when it came to deflating cuff pressure and determination of
diastolic pressure. The difference of the diastolic errors can be easily explained: the
cardiac patient group exhibited a slower pressure pulse slope and therefore a larger delay
time change during the measurement, yielding a clear indication of the changing point.
In the EP method, when cuff pressure exceeds the venous pressure level, the volume
of blood in the hand starts to increase with each heart beat. This induces vascular
congestion and a pressure increase in the veins and surrounding tissues on the distal side
of the cuff. The velocity of the pressure pulse waves and therefore also the delay change
depends exponentially on vascular compliance and the pressure difference between the
inner and outer arterial wall. A change in this pressure, also known as transmural
pressure, reduces pulse wave velocity and slightly increases pulse delay, before the
diastolic blood pressure level is achieved. This phenomenon is illustrated in Fig. 28.
Intra-arterial (IA) blood pressure is measured invasively with a catheter tip and radial
artery pulsation noninvasively with an electronic palpation transducer. As seen, the
recorded venous pressure increases slightly until cuff pressure drops under it. The same
69
100
0
0
-100
-200
TimeTime
delay (ms)
delay (ms)
-200
50
5
10
5
15
10
15
20
25
Time (s)
20
Time (s)
30
25
35
30
40
35
40
Electronic palpation signal (a.u.)
200 200
Electronic palpation signal
IA & IV pressure (mmHg)
Cuff pressure, intra-arterial
and intra-venus
pressure (mmHg)
Cuff pressure,
phenomenon occurs also with inflating cuff pressure. Owing to venous congestion, the
pulse is still slightly delayed at intermediate cuff pressures between the diastolic and
intravenous level. The vertical line in the centre represents the average diastolic blood
pressure value during the measurement, and the other lines its minimum and maximum
values. According to our experiments (Papers VI, VII & VIII), the deflating mode is
preferable, due to a smaller increase in vascular pressure. In addition, patients might have
had a larger capillary and vein volume on the distal side of the cuff, so the pressure
increase in the hand during cuff deflation was smaller than the corresponding value for
the healthy volunteers.
50
0
0
-50
0
-50
0
20
20
40
40
60
80
100
Cuff pressure (mmHg)
60
80
Cuff pressure (mmHg)
100
120
140
120
140
Fig. 28. Upper figure: pulsations in right (intra vascular, IV) and the left (noninvasive
electronic palpation sensor) radial artery, cuff pressure and intra vascular (IV) venous
pressure as a function of time. Lower figure: pulse delay between pulses as a function of cuff
pressure. The vertical lines in the lower figure are the minimum, average and maximum
values of IA diastolic blood pressure. (VI)
On other hand, the inadequate accuracy observed in systolic blood pressure determination
with the healthy group was unexpected, but may be due to imprecise system calibration; a
level change of ten centimetres yields an error of nearly 8 mmHg, owing to hydraulic
pressure in the artery. The author did not attend the tests with the healthy volunteers.
In the EP method, the ECG signal, being easily detectable in normal subjects, can be
used as a timing reference. However, since electrical cardiac activation (ECG) does not
constitute mechanical activation (Hong 2000), it is advisable to use pressure pulses
70
instead. The R-spike is followed by a pre-ejection period before the aortic valve opens.
This time period depends on aortic diastolic pressure and heart constriction rate, and the
resulting deviation in pressure pulse delay renders the determination of blood pressure
inaccurate (Meigas et al. 2001). The optimal reference point is the radial artery of the
other hand, which usually produces an almost identical pulse waveform, thereby greatly
facilitating the accurate detection of pulse delays. This method does not provide any
information on pulse wave velocity, which could be used for continuous, pulse-to-pulse
monitoring of blood pressure (Chen et al. 2000, Meigas et al. 2001 and 2003).
Nevertheless, pulse wave velocity can be clearly specified in measurements made on one
hand using, for example, the brachial and radial arteries.
The shape of pressure pulses, especially optically measured plethysmographic pulses,
is quite smooth, and does not provide a very distinct timing point. A better timing
reference may be obtained with optical interferometers, such as laser-Doppler systems
based on the self-mixing effect (Hast et al. 2001-2003, Hast 2003, paper IX and Meigas
et al. 2001 and 2003). These interferometers measure skin vibrations over an artery with
a very high degree of accuracy, restricted theoretically only by the wavelength of the laser
divided by two (~300 - 700 nm). Their main advantage is the absence of a loading effect,
as the sensor is not in contact with the skin. Also, unlike tonometer type pressure
transducers (Drzewiecki et al. 1983 and 1996, Kelly et al. 1989, Kemmotsu et al. 1991,
Bansal et al. 1994, Drzewiecki and Pilla 1998, Fetics et al. 1999), laser-Doppler
interferometers do not suffer from sensor movements caused by arterial pulsations.
Further, their timing accuracy is very good, because the signal has a much wider
bandwidth than a pressure signal.
In principle, the time delay method can be used with any transducer that measures
pulsations either on the radial artery or a finger. One possibility is to measure
noninvasively radial artery pressure pulses in both hands using either two identical
pressure array transducers or interferometers with the same bandwidth. In addition, by
measuring the starting time of pressure pulses (opening of the aortic valve) using, for
example, a contact microphone on the chest, we will be able to determine the pulse
waveform and pulse wave velocity (PWV), which enables noninvasive, pulse-by-pulse
monitoring of blood pressure. This requires a daily calibration measurement using an arm
cuff. Also pressure pulses measured in the brachial artery under the cuff can be used as a
timing base. Finally, detecting the waveform, amplitude and time of pulsations in the
radial artery can be utilized in many ways: for example, with other noninvasively
measured physiological signals, they can be used to determine arterial compliance or
stiffness (O’Rourke et al. 1992, Asmar et al. 1995, Safar et al. 1998, Blacher et al. 1999,
O’Rourke and Mancia 1999 and Asmar 1999) and central aortic pressure (Kelly et al.
1989, Karamonoglu et al. 1993, Chen et al. 1997 and Fetics et al. 1999).
The mean errors obtained with the OSC method in the cardiac patient group were too
large to meet the AAMI and BHS standards, which call for an inaccuracy level of less
than 5 mmHg and a standard deviation of less than 8 mmHg. Our analysis employed a
scaling multiplier to get the 95% level of the oscillation curve to cross at the mean intraarterial pressure point. Without a scaling multiplier, the maximum oscillation in Fig. 26 is
about 78 mmHg, which overestimates mean blood pressure by 13 mmHg. The
corresponding figures for the EP method in the same measurement are 56/94 mmHg for
diastolic and systolic pressure, while intra-arterial diastolic and systolic pressures
71
alternate between 47 and 54 mmHg and 89 and 99 mmHg, respectively. This shows that
the point where the change of delay appears gives a reasonably good estimation of
diastolic blood pressure. In the OSC method, however, maximum oscillation amplitude is
attained at a pressure higher than mean intra-arterial pressure, producing false readings.
6 Conclusion
This study first described noninvasive blood pressure measurements in detail, starting
from the historical background and extending to modern methods.
The next stage involved designing, building and testing EMFi-based sensor arrays,
which were subsequently found to be adequate for monitoring blood pressure pulsations
in the radial artery. Their material properties were appropriate for long-term low-power
applications based on sensing force/pressure pulsations at near room temperature.
Nonetheless, the material’s unhomogeneity still obstructs their application to blood
pressure measurement systems of the tonometer type.
At the following stage, the developed measurement system was used at the Oulu
University Hospital for two sets of measurements, conducted in 1997 – 1999. In the
second set, the measurement system consisted of a microcontroller-based test device for
determining blood pressure values and a data recording measurement system for further
analysis. The test device worked as intended.
The determination of diastolic blood pressure was described, and the available
methods’ accuracy was studied. Using intra-arterial blood pressure as a reference, this
study investigated two measuring methods, the oscillometric (OSC) and the electronic
palpation (EP) method. The test subjects comprised a group of healthy volunteers and a
group of cardiac patients, who had undergone either a bypass or a valve operation, or
both. The healthy group was measured twice using both inflating and deflating cuff
pressure in a sitting and supine position with a one-minute follow-up period between the
measurements. The cardiac patient group, on the other hand, was measured only in the
supine position with increasing cuff pressure. All measurements were analyzed with the
aim of ensuring the best possible accuracy level for both groups. Thus, in the OSC
method, case-specific characteristic ratios were used to reach an ideal degree of accuracy.
In the EP method, the determination of systolic blood pressure was based on the last
palpated pulse, while the determination of diastolic pressure relied on changes in the time
delay of the detected signals produced by the occlusion cuff. For better timing accuracy,
interpolation was used to increase the sampling rate by a factor of ten.
In the oscillometric method, the acquired characteristic ratios for the determination of
diastolic blood pressure differed considerably from those reported in the literature,
whereas the characteristic ratios for systolic blood pressure were within the reported
73
ranges. It was also established that cuff pressure mode affects accuracy: in the deflating
mode, the characteristic ratios were slightly higher than in the inflating mode. Also
posture had an effect: both characteristic ratios increased by about 4-8 percent when the
test subjects were sitting. Age, on the other hand, seemed to have only a minute effect on
these characteristic ratios, provided that the same cuff pressure mode and position were
used in all measurements.
In the healthy group, the EP method was found to provide a slightly better degree of
accuracy with deflating cuff pressure. With a mean error of about zero and a standard
deviation of about 5-6 mmHg, also the inflating mode used in the cardiac patient group
gave excellent results. Nevertheless, to avoid vascular swelling, the deflating pressure is
the more recommendable of the two. To sum up, a comparison between the OSC and the
EP method indicates that the EP method offers a feasible alternative to the OSC method,
particularly because it also produces information on the delay and waveform of pressure
pulses. As a consequence, it can be applied to continuous blood pressure monitoring and
the noninvasive determination of such physiological indices as arterial compliance and
central aortic pressure.
References
Alametsä J, Koivuluoma M & Varri A (2001) Mobile physiological signal measurement station.
Proc. International Federation for Medical and Biological Engineering, 9th Mediterranean
Conference on Medical and Biological Engineering and Computing, MEDICON, Pula, Croatia
1: 289–292.
Alametsä J, Värri A, Koivuluoma M & Barna L (2004) The potential of EMFi sensors in heart
activity monitoring. 2nd OpenECG Workshop Integration of the ECG into the HER &
interoperability of ECG device systems, Berlin, Germany.
Junnila S Akhbardeh A, Varri A & Koivistoinen T (2005) An EMFi-film sensor based
ballistocardiographic chair: performance and cycle extraction method, Signal Processing
Systems Design and Implementation, IEEE Workshop on, Athens, Greece 1: 373–377.
Akhbardeh A Junnila S Koivistoinen T Koobi T & Varri A (2005) Ballistocardiogram classification
using a novel transform so-called AliMap and biorthogonal wavelets. Intelligent Signal
Processing, IEEE International Workshop 1: 64–69.
Koivistoinen T, Junnila S, Varri A & Koobi T (2004) A new method for measuring the
ballistocardiogram using EMFi sensors in a normal chair. EMBC Proc. 26th Annual
International Conference 1(3): 2026–2029.
Anttonen J & Surakka V (2005) Emotions and heart rate while sitting on a chair. CHI2005,
Portland, Oregon, USA: 491–499.
Alexander H, Cohen M & Steinfeld L (1977) Criteria in the choice of an occluding cuff for the
indirect measurement of the blood pressure. Med Biol Eng Comput 15: 2–10.
Allen J, Gehrke T, O’Sullivan JJ, King ST & Murray A (2004) Characterization of the Korotkoff
sounds using joint time-frequency analysis. Physiol Meas 25: 107–117.
Antila M, Muurinen T, Linjama J & Nykänen H (1997) Measurement methods of flat panel
electromechanical film loudspeakers. Proc. Active 97, Budabest, Hungary 1: 607–618.
Antila M, Saarinen K & Kirjavainen K (1998) Performance of a large, flat electromechanical film
actuator panel. Proc. Actuator 98, 6th International Conference on New Actuators 1: 96–99.
Asada H, Reisner A, Shaltis P & McCombie D (2005) Towards the Development of Wearable
Blood Pressure Sensors: A Photo-Plethysmograph Approach Using Conducting Polymer
Actuators. IEEE-EMBS, Proc. 27th Annual International Conference, Shanghai, China: 4156–
4159.
Asada HH, Shaltis P, Reisner A Rhee S & Hutchinson RC (2003) Mobile monitoring with wearable
photoplethysmographic biosensors. IEEE Eng Med and Biol Magazine 22(3): 28–40.
Asmar R (1999) Arterial Stiffness and Pulse Wave Velocity, Clinical Applications. Éditions
scientifiques et médicales Elsevier SAS. 23, rue Linois, 75724 Paris cedex 15.
75
Asmar R, Benetos A, Topouchian J, Laurent P, Pannier B, Brisac A-M, Target R & Levy BI (1995)
Assessment of arterial distensibility by automatic pulse wave velocity measurement.
Hypertension 26(3): 485–490.
Association for the Advancement of Medical Instrumentation. American national standard.
Electronic or automated sphygmomanometers. ANSI/AAMI SP 10–1987. Arlington, VA:
AAMI, 1987: 25.
Association for the Advancement of Medical Instrumentation. American national standard.
Electronic or automated sphygmomanometers. ANSI/AAMI SP 10–1992. Arlington, VA:
AAMI, 1993: 1–40.
Backman J & Karjalainen M (1990) Audio and ultrasonic transducers based on electrothermomechanical film (ETMF). Proc. IEEE Int. Conf. on acoustics, speech and signal processing,
Albuquerque, NM, USA 2: 1173–1176.
Backman J (1990) Audio applications of electrothermomechanical film (ETMF). J Aud Eng Soc
38: 364–371.
Bansal V, Drzewiecki G & Butterfield R (1994) Design of a flexible diaphragm tonometer, Proc.
13th S Bioeng Conf, Washington, DC, USA: 148–151.
Birch AA & Morris SL (2003) Do the FinapresTM and Colin® radial artery tonometer measure the
same blood pressure changes following deflation of thigh cuffs? Physiol Meas 24(3): 653–660.
Blacher J, Asmar R, Djane S, London GM & Safar ME (1999) Aortic pulse wave velocity as a
marker of cardiovascular risk in hypertensive patients. Hypertension 33 (5): 1111–1117.
Boehmer RD (1987) Continuous, real-time, noninvasive monitor of blood pressure: Penaz
methodology applied to the finger. J Clin Monit 3(4): 282–287.
Bramwell JC & Hill AV (1922) The velocity of the pulse wave in man. Proc. Royal Soc Biom, 96:
298–306.
Bur A, Hirschl MM, Herkner H, Oschatz E, Kofler J, Woisetschlager C & Laggner AN (2000)
Accuracy of oscillometric blood pressure measurement according to the relation between cuff
size and upper-arm circumference in critically ill patients. Crit Care Med 28(2): 371–376.
Callaghan FJ, Geddes LA, Babbs CF & Bourland JD (1986) Relationship between pulse-wave
velocity and arterial elasticity. Med Biol Eng Comput: 248–254.
Chan KW, Hung K & Zhang YT (2001) Noninvasive and cuffless measurements of blood pressure
for telemedicine. IEEE-EMBS, Proc. 23rd Annual International Conference, Istanbul, Turkey
(4): 3592–3593.
Chen C-H, Nevo E, Fetics B, Pak PH, Yin FC, Maughan WL & Kass DA (1997) Estimation of
central aortic pressure waveform by mathematical transformation of radial tonometry pressure;
validation of generalized transfer function. Circulation 95(7): 1827–1836.
Chen W, Kobayashi T, Ichikawa S, Takeuchi Y & Togawa T (2000) Continuous estimation of
systolic blood pressure using the pulse arrival time and intermittent calibration. Med Biol Eng
Comput 38(5): 569–574.
Ciaccio E & Drzewiecki G (1988) Array sensor for arterial pulse recording-reduction of motion
artifact. Bioengineering Conference, Proc. 14th Annual Northeast, Durham, NH, USA: 66–69.
Ciaccio E, Drzewiecki G & Karam E (1989) Algorithm for reduction of mechanical noise in arterial
pulse recording with tonometry. Bioengineering Conference, Proc. 15th Annual Northeast: 161–
162.
Colak S & Isik C (2003) Fuzzy oscillometric blood pressure classification. Fuzzy Information
Processing Society, NAFIPS, 22nd International Conference of the North American, USA: 208–
213.
Colak S & Isik C (2003) Systolic blood pressure classification. Neural Networks, Proc.
International Joint Conference on 1: 627–630.
Colak S & Isik C (2004) Blood pressure estimation using neural networks. Proc. IEEE International
Conference on Computational Intelligence for Measurement Systems and Applications: 21–25.
76
Colak S & Isik C (2004) Evaluation of pulse features for oscillometry. Proc. IEEE 30th Annual
Northeast Bioengineering Conference: 93–94.
Colak S & Isik C (2004) Simulation of oscillometric waveforms. Canadian Conference on
Electrical and Computer Engineering: 497–500.
Cozby RR & Adhami RR (1993) Low-Frequency Korotkoff signal analysis and application. IEEE
Trans Biomed Eng 40(10): 1067–1070.
Cristalli C, Neuman MR & Ursino M (1994) Studies on soft tissue pressure distribution in the arm
during non-invasive blood pressure measurement. Engineering Advances: New Opportunities
for Biomedical Engineers, Proc. 16th Annual International Conference of the IEEE-EMBS 1:
41–42.
Cristalli C, Mancini P & Ursino M (1992) An experimental and mathematical study of noninvasive
blood pressure measurement. Proc. Annual International Conference of the IEEE-EMBS 14(1):
63–64.
Cristalli C, Ursino M, Palagi F & Bedini R (1993) FEM simulation and experimental evaluation of
the 'squeezing' phenomenon in Riva-Rocci blood pressure measurement. Proc. Comput
Cardiology, London, Great Britain: 655–658.
Drzewiecki G & Pilla JJ (1998) Noninvasive measurement of the human brachial artery pressurearea relation in collapse and in hypertension. Ann Biomed Eng 26: 967–974.
Drzewiecki G (1995) Noninvasive assessment of arterial blood pressure and mechanics. In:
Bronzino JD (ed) The Biomedical Engineering Handbook. Trinity College, Hartford,
Connecticut, USA: 1198.
Drzewiecki G, Bansal V, Karam E, Hood R & Apple A (1993) Mechanics of the occlusive arm cuff
and its application as a volume sensor. IEEE Trans Biomed Eng 40: 704–708.
Drzewiecki G, Hood R & Apple A (1994) Theory of the oscillometric maximum and the systolic
and diastolic detection ratios. Ann Biomed Eng 22: 88–96.
Drzewiecki G, J Melbin & Noordengraaf A (1983) Arterial tonometry: review and analysis. J
Biomech 16: 141–152.
Drzewiecki G, Solanki B, Wang J-J & Li JK-L (1996) Noninvasive determination of arterial
pressure and volume using tonometry. ELECTRO '96. Professional Program. Proc 1: 61–63.
Epstein RH, Huffnagle S & Bartkowski RR (1991) Comparative accuracies of a finger blood
pressure monitor and an oscillometric blood pressure monitor. J Clin Monit 7(2): 161–167.
Epstein RH, Kaplan S, Leighton BL, Norris MC & DeSimone CA (1989) Evaluation of a
continuous noninvasive blood pressure monitor in obstetric patients undergoing spinal
anesthesia. J Clin Monit 5(3): 157–163.
Erlanger JA (1904) A new instrument for determining the minimum and maximum blood-pressure
in man. Johns Hopkings Hosp Rep 12: 53–110.
Fetics B, Nevo E, Chen C-H & Kass D (1999) Parametric model derivation of transfer function for
noninvasive estimation of aortic pressure by radial tonometry. IEEE Trans Biomed Eng 46(6):
698–706.
Ficher M, Kirjavainen K, Vainikainen P & Nyfors E (1990) Sensor for the measurement of
pressure. Proc. 20th European microwave conference, Budapest, Hungary 1: 985–989.
Franchi D, Bedini R, Manfredini F, Berti S, Palagi G., Ghione S & Ripoli A (1996) Blood pressure
evaluation based on arterial pulse wave velocity. Computers in Cardiology 1: 397–400.
Friedman DB, Jensen FB, Matzen S & Secher NH (1990) Non-invasive blood pressure monitoring
during head-up tilt using the Penaz principle. Acta Anaesthesiol Scand 34(7): 519–522.
Friesen RH & Lichter IL (1981) Indirect measurement of blood pressure in neonates and infants
utilizing an automatic noninvasive oscillometric monitor. Anesth Analg 10: 742–745.
Geddes LA & Whistler SJ (1978) The error in indirect blood pressure measurement with incorrect
size of cuff. Am Heart J 96: 4.
77
Geddes LA, Boelz MH, Babbs CF, Bourland JD & Tacker WA (1981) Pulse transit time as a
indicator of arterial blood pressure. Psychophysiology 18: 71–74.
Geddes LA, Voelz M, James S & Reiner D (1981) Pulse wave velocity as a method of obtaining
systolic and diastolic blood pressure. Med Biol Eng Comput 19: 671–672.
Geddes LA, Voelz M & Combs C (1983) Characterization of the oscillometric method for
measuring indirect blood pressure. Ann Biomed Eng 10: 271–280.
Geddes LA (1991) Handbook of Blood Pressure Measurement. Humana Press.
Gerhard-Multhaupt R (2002) Less can be more: Holes in polymers lead to a new paradigm of
piezoelectric materials for electret transducers. IEEE Trans Dielectics and Electrical Insulation
9: 850–859.
Gerhard-Multhaupt R (2002) Voided polymer electrets –New materials, new challenges, new
changes. Proc. 11th International Symposium on Electrets: 36–45.
Gerhard-Multhaupt R, Wegener M, Wirges W, Giacometti JA, Altafilm RAC, Santos LF, Faria RM
& Paajanen M (2002) Electrode poling of cellular polypropylene films with short high-voltage
pulses. Proc. Annual Report Conference on Electrical Insulation and Dielectric Phenomena:
299–302.
Giller CA & Aaslid R (1994) Estimates of pulse wave velocity and measurement of pulse transit
time in the human cerebral circulation. Ultrasound Med Biol 20(2): 101–105.
Gravenstein JS, Paulus DA, Feldman J & McLaughlin G (1985) Tissue hypoxia distal to a Peñaz
finger blood pressure cuff. J Clin Monit 1(2): 120–5.
Hämäläinen MK, Parviainen JK & Jääskeläinen T (1996) A novel micromovement actuator
manufactured using plastic electromechanical film. Review of Scientific Instruments 67: 1598–
1601.
Hasegawa H, Ozawa M, Kanai H, Hoshimiya N & Chubashi N (1997) Evaluation of the elastic
property of the arterial wall by measuring small velocity signals using ultrasound. IEEE
Ultrasonics Symposium: 1169–1172.
Hast J T (2003) Self-Mixing Interferometry and Its Applications in Noninvasive Pulse Detection.
Oulu, Finland. Doctoral thesis, University of Oulu, Department of Electrical Engineering.
Hast J, Myllylä R, Sorvoja H & Miettinen J (2001) A noninvasive laser Doppler system to study
cardiovascular pulsation and its properties. Asian J Phys 10(4): 15 p.
Hast J, Myllylä R, Sorvoja H & Miettinen J (2001) Self-mixing interferometry in noninvasive pulse
wave velocity measurement. Molecular and Quantum Acoustics 22: 95–106.
Hast J, Myllylä R, Sorvoja H & Miettinen J (2002) Comparison between elasticity of the arterial
wall and pulse wave velocity with laser Doppler method. Proc SPIE 4619: 259–268.
Hast J, Myllylä R, Sorvoja H & Miettinen J (2003) Arterial pulse shape measurement using selfmixing interferometry. Proc. International Society for Optical Engineering (SPIE), Coherence
domain optical methods in biomedical science and clinical applications VII, San Jose, USA,
4956: 341–347.
Heikkinen LM, Panula HE, Lyyra T, Olkkonen H, Kiviranta I, Nevalainen T & Helminen HJ
(1997) Electromechanical film sensor device for dynamic force recordings from canine limbs.
Scan J Lab Animal Science 24: 85–92.
Hong H (1994) Optical Interferometric Measurement of Skin Vibration for the Diagnosis of
Cardiovascular Diseases. Michican: UMI Dissertation Services, A Bell & Howell Company:
111–113.
Hori C, Itakura K, Nogawa M, Shirakabe M, Kubota I, Tomoike H & Takatani S (1997) Estimation
of aortic BP waveform from noninvasive radial tonometry; validation of FFT and ARX
methods. Proc. 19th Annual International Conference of the IEEE-EMBS 3: 1142–1145.
Howell WH & Brush CE (1901) A critical note upon clinical methods of measuring blood pressure.
Boston Med Surg J 145: 146–151.
78
Hung K, Zhang YT & Tai B (2004) Wearable medical devices for tele-home healthcare. Proc. 26th
Annual International Conference of the IEEE-EMBS, San Francisco, USA 2: 5384–5387.
Imholz BPM, Wieling W, van Montfrans GA & Wesseling KH (1998) Fifteen years experience
with finger arterial pressure monitoring: assessment of the technology. Cardiovascular Research
38(3): 605–616.
Jagomägi K, Raamat R & Talts J (2001) Effect of altering vasoactivity on the measurement of
finger blood pressure. Blood Press Monitor 6(1): 33–40.
Jagomägi K, Talts J, Raamat R & Lansimies E (1996) Continuous non-invasive measurement of
mean blood pressure in fingers by volume-clamp and differential oscillometric method. Clin
Physiol 16(5): 551–560.
Kanai H, Kawabe K, Takano M, Murata R, Chubashi N & Koiwa Y (1994) New method for
evaluating local pulse wave velocity by measuring vibrations on arterial wall. Electronics letters
30(7): 534–536.
Kanai H, Satoh H, Hirose K & Chubashi N (1993) A new method for measuring small local
vibrations in the heart using ultrasound. IEEE Trans Biomed Eng 40(12): 1233–1242.
Karamanoglu M, O’Rourke MF, Avolio AP & Kelly RP (1993) An analysis of the relationship
between central aortic and peripheral upper limb pressure waves in man. E Heart J 14: 160–167.
Kawabe K, Kanai H & Chubashi N (1993) Accuracy evaluation in ultrasonic-Doppler-based
measurement of small vibrations for acoustical diagnosis of the aortic wall. Electronics Letters
29(10): 915–916.
Kelly R, Hayward C, Ganis J, Daley J, Avolio A & O’Rourke M (1989) Noninvasive registration of
the arterial pressure pulse waveform using high-fidelity applanation tonometry. J Vasc Med
Biol 1: 142–149.
Kemmotsu O, Ueda M, Otsuka H, Yamamura T, Winter DC & Eckerle JS (1991) Arterial
tonometry for non-invasive, continuous blood pressure monitoring during anaesthesia.
Anaesthesiology 75(2): 333–340.
Kerola J, Kontra V & Sepponen R (1997) Non-invasive blood pressure data acquisition employing
pulse transit time detection. Proc. 18th Annual International Conference of the IEEE-EMBS,
Amsterdam, Netherlands: 1308–1309.
Kimble KJ, Darnall RA & Yelderman M et al. (1981) An automatic oscillometric technique for
estimating mean arterial pressure in clinically ill neonates. Anesthesiology 54(5): 423–5.
Kirjavainen K (1987) Electromechanical film and procedure for manufacturing same. U.S. Patent
4,654,546.
Korteweg DJ (1878) Uber die Fortpflanzungsgeshwindkeit des Schalles in elastichen Röhren. Ann
Phys Chem 5: 525–542.
Kugler J, Rollnik J & Schmitz N (1997) Retest-reliability and convergent validity of noninvasive
blood pressure determination: arm sphygmomanometry vs. Penaz-method. Int J Clin Monit
Comput 14(4): 251–254.
Kurki T, Smith NT, Head N, Dec-Silver H & Quinn A (1987) Noninvasive continuous blood
pressure measurement from the finger: optimal measurement conditions and factors affecting
reliability. J Clin Monit 3(1): 6–13.
Kurki TS, Smith NT, Sanford TJ & Head N (1989) Pulse oximetry and finger blood pressure
measurement during open-heart surgery. J Clin Monit 5(4): 221–228.
Langewouters GJ, Settels JJ, Roelandt R & Wesseling H (1998) Why use Finapres or Portapres
rather than intra-arterial or intermittent non-invasive techniques of blood pressure
measurement? J Med Eng Technol 22(1): 37–43.
Lee CM & Zhang YT (2003) Cuffless and noninvasive estimation of blood pressure based on a
wavelet transform approach. Biomedical Engineering, IEEE-EMBS Asian-Pacific Conference:
148–149.
79
Lee CM & Zhang YT (2003) Reduction of motion artifacts from photoplethysmographic recordings
using a wavelet denoising approach. Biomedical Engineering, IEEE-EMBS Asian-Pacific
Conference: 194–195.
Lekkala J & Paajanen M (1999) EMFi–new electret material for sensors and actuators. Proc. 10th
International Symposium of Electrets: 743–746.
Lekkala J (1997) The sound of silence… and other applications of the electromechanical film
technology. Industrial Horizons: 12–13.
Lekkala J, Poramo R, Nyholm K & Kaikkonen T (1996) EMF force sensor–a flexible and sensitive
electret film for physiological applications. Med Biol Eng Comput 34(1): 67–68.
Lin C-T, Liu S-H, Wang J-J & Wen Z-W (2003) Reduction of interference in oscillometric arterial
blood pressure measurement using fuzzy logic. IEEE Trans Biomed Eng 50(4): 432–441.
Lindener M, Bauer-Gogonea S, Bauer S, Paajanen M & J Raukola J (2002) Dielectric barrier
microdischarges: Mechanisms for the charging of cellular piezoelectric polymers. J Appl Phys
91(8): 5283–5287.
Liu S-H & Lin C-T (2001) A model-based fuzzy logic controller with Kalman filtering for tracking
mean arterial pressure. IEEE Trans Systems Man Cyber Part A 31(6): 676–686.
Maestri R, Pinna GD, Robbi E, Capomolla S & La Rovere MT (2005) Noninvasive measurement
of blood pressure variability: accuracy of the Finometer monitor and comparison with the
Finapres device. Physiol Meas 26: 1125–1136.
Maguire M, Ward T, Markham C, O’Shea D & Kevin L (2001) A comparative study in the use of
brachial photoplethysmography and the QRS complex as timing reference in determination of
pulse transit time. Proc. 23rd Annual International Conference of the IEEE-EMBS, Istanbul,
Turkey: 215–218.
Marey EJ (1876) Physiology Experimentale. Paris, Masson & Cie.
McDonald DA (1974) Blood flow in arteries. London, Eward Arnold, 2nd edition, Reading 253–
254.
Meigas K, Hinrikus H, Kattai R & Lass J (2003) Self-mixing in a diode laser as a method for
cardiovascular diagnostics. J Biomed Optics 8(1): 152–160.
Meigas K, Kattai R & Lass J (2001) Continuous blood pressure monitoring using pulse wave delay.
Proc. 23rd Annual International Conference of the IEEE-EMBS 4: 3171–3174.
Moens AI (1878) Die Pulscurve. E J Brill, Leiden 1: 87–95.
Mohiaddin RH, Firmin DN & Longmore DB (1993) Age-related changes of human aortic flow
wave velocity measured noninvasively by magnetic resonance imaging. J Appl Phys 74: 492–
497.
Molhoek GP, Wesseling KH, Settels JJ, van Vollenhoven E, Weeda HW, de Wit B & Arntzenius
AC (1984) Evaluation of the Penaz servo-plethysmo-manometer for the continuous, noninvasive measurement of finger blood pressure. Basic Res Cardiol 79: 598–609.
Moraes JCTB & Cerulli M (1999) Development of a new oscillometric blood pressure
measurement system. Proc. Comp in Card, 1: 467–470.
Moraes JCTB & Cerulli M (2000) A strategy for determination of systolic, mean and diastolic
blood pressures from oscillometric pulse profiles. Proc Com in Cardiol 1: 211–214.
Narimatsu K & Ohmori K (1992) An algorithm for tonometric absolute blood pressure
measurement. Proc. Computers in Cardiology, Durham, NC, USA: 559–562.
Nelesen RA & Dimsdale JE (2002) Use of radial arterial tonometric continuous blood pressure
measurement in cardiovascular reactivity studies. Blood Press Monitor 7(5): 259–263.
Neugschwandtner G, Schwödiauer R, Bauer-Gogonea S, Bauer S, Paajanen M & Lekkala J (2001)
Piezo-and pyroelectricity of a polymer-foam space-charge electret. J Appl Phys 89(8): 4503–
4511.
Nissilä S & Miettinen J (2000) Measuring arrangement. European patent application EP 1 034 736
A1.
80
Novak V, Novak P & Schondorf R (1994) Accuracy of beat-to-beat noninvasive measurement of
finger arterial pressure using the Finapres: A spectral analysis approach. J Clin Monit 10(2):
118–126.
Nykänen H, Antila M, Kataja J, Lekkala J & Uosukainen S (1999) Active control of sound based
on utilizing EMFi technology. Proc. Active 99, Fort Lauderdale: 1159–1170.
O’Brien E, Petrie J, Littler W et al. (1990) The British Hypertension Society protocol for the
evaluation of automated and semi-automated blood pressure measuring device with special
reference to ambulatory systems. J Hypertension 8: 607–619.
O’Brien, E, Petrie J, Littler W et al. (1993) Short report: an outline of the revised British
Hypertension Society protocol for the evaluation blood pressure measuring devices. J
Hypertension 11: 677–679.
O'Brien E, Petrie J, Littler W, de Swiet M, Padfield PL, Altman DG, Bland M, Coats A, Atkins N
(1993) An outline of the revised British Hypertension Society protocol for the evaluation of
blood pressure measuring devices. J Hypertens 11: 677–679.
O’Brien E, Waeber B, Parati G, Staessen J & Myers MG (2001) Clinical Review: Blood pressure
measuring devices: recommendations of the European Society of Hypertension. BMJ 322: 531–
536.
O'Brien E, Pickering T, Asmar R, Myers M, Parati G, Staessen J, Mengden T, Imai Y, Waeber B,
Palatini P with the statistical assistance of Atkins N & Gerin W on behalf of the Working Group
on Blood Pressure Monitoring of the European Society of Hypertension. (2002) International
protocol for validation of blood pressure measuring devices in adults. Blood Press Monitor 7: 3–
17.
O’Brien E, Asmar R, Beilin L, Imai Y, Mancia G, Mengden T, Myers M, Padfield P, Palatini P,
Parati G, Pickering T, Redon J, Staessen J, Stergiou G, Verdecchia P on behalf of the European
Society of Hypertension Working Group on Blood Pressure Monitoring. (2005) Practice
guidelines of the European Society of Hypertension for clinic, ambulatory and self blood
pressure measurement. J Hypertension 23: 697–701.
O’Rourke M & Mancia G (1999) Arterial Stiffness. J Hypertension. 17(1): 1–4.
O’Rourke M, Kelly R & Avolio A (1992) The Arterial Pulse. 200 Chester Field Parkway, Malvern,
Pennsylvania 19355–9725, USA., Lea & Febiger.
Okada M (1988) Possible determinants of pulse-wave velocity in vivo. IEEE Trans Biomed Eng
35(5): 357–361.
Paajanen M (2001) The cellular polypropylene electret material-Electromechanical properties. D.
Tech. Doctoral Thesis, Department of Electrical Engineering, Tampere University of
Technology, VTT Publications, Technical Research Center of Finland, Espoo, Vol. 436.
Paajanen M, Lekkala J & Kirjavainen K (2000) Electromechanical film (EMFi) –a new
multipurpose electret material. Sensors and Actuators 84: 95–102.
Paajanen M, Lekkala J & Välimäki H (2001) Electromechanical modeling and properties of the
electret film EMFI. IEEE Trans Dielectrics and Electrical Insulation 8(4): 629–636.
Paajanen M, Minkkinen H & Raukola J (2002) Gas diffusion expansion –Increased thickness and
enhanced electromechanical response of cellular polymer electret films. Proc. 11th International
Symposium on Electrets: 191–194.
Paajanen M, Välimäki H & Lekkala J (1999) Modelling the sensor and actuator operations of the
electromechanical film EMFi. Proc. 10th International Symposium of Electrets: 735–738.
Paajanen M, Välimäki H & Lekkala J (2000) Modelling the electromechanical film (EMFi). J
Electrostatics 48: 193–204.
Paajanen M, Wegener M & Gerhard-Multhaupt R (2001) Charging of the cellular space-charge
electret films in various gas atmospheres. Proc. Annual Report Conference on Electrical
Insulation and Dielectric Phenomena: 24–27.
81
Paajanen M, Wegener M & R Gerhard-Multhaupt (2001) Understanding the role of the gas in the
voids during corona charging of cellular electret films –a way to enhance their piezoelectricity. J
Phys D: Appl Phys 34: 2482–2488.
Pachon V (1909) Oscillometre sphygmomanometrique a grande sensibilite constante. Comptes
Rendus Soc Biol 66: 776–779.
Peltonen J, Paajanen M & Lekkala J (2000) Determination of the actuator sensitivity of
electromechanical polypropylene films by atomic force. J Appl Phys 88(8): 4789–4793.
Penaz J (1973) Photoelectric measurement of blood pressure, volume and flow in the finger. Proc.
Digest of the 10th International Conf. EMBS, Dresden, Germany: 104.
Penaz J (1992) Criteria for set point estimation in the volume clamp method of blood pressure
measurement. Physiol Res 41(1): 5–10.
Penzel T, Brandenburg U, Peter J H & Stamnitz A (1993) Portable continuous non-invasive blood
pressure recording for sleep studies. Proc. 15th Annual International Conference of the IEEEEMBS, San Diego, CA, USA: 393–394.
Pirttikangas S, Suutala J Riekki J & Röning J (2004) Learning vector quantization in footstep
identification. Proc. Pervasive 2004, Second International Conference on Pervasive Computing.
Poon CCY & Zhang YT (2005) Cuff-less and Noninvasive Measurements of Arterial Blood
Pressure by Pulse Transit Time IEEE-EMBS, Proc. 27th Annual International Conference,
Shanghai, China 1: 5877–5880.
Posey JA, Geddes LA, Williams H & Moore AG (1969) The meaning of the point of maximum
oscillation in cuff pressure in the indirect measurement of blood pressure. Part 1. Cardiovasc
Res Cent Bull 8(1): 15–25.
Pressman G & Newgard P (1963) A transducer for continuous external measurement of arterial
blood pressure. IEEE Trans Biomed Electonics 10: 73–81.
Raamat R, Jagomägi K & Talts J (2000) Different responses of Finapres and the oscillometric
finger blood pressure monitor during intensive vasomotion. J Med Eng Technol 24(3): 95–101.
Raamat R, Jagomägi K, Talts J, Lansimies E, Jurvelin J & Kolari P (2001) Continuous mean
arterial pressure measurement in the fingers: the influence of local arm cooling. Med Biol Eng
Comput 39(5): 584–589.
Raamat R, Jagomägi K, Talts J, Toska K & Walloe L (2001) Beat-to-beat measurement of the
finger arterial pressure pulse shape index at rest and during exercise. Clin Physiol Funct
Imaging 23(2): 87–91.
Raamat R, Jagomägi K, Talts J, Toska K & Walloe L (2001) Recording of short-term finger blood
pressure changes induced by an arterial occlusive thigh cuff: comparison between the modified
oscillometric and Finapres techniques. Physiol Meas 22(2): N13–20.
Raamat R, Talts J, Jagomägi K & Lansimies E (1999) Mathematical modelling of non-invasive
oscillometric finger mean blood pressure measurement by maximum oscillation criterion. Med
Biol Eng Comput 37(6): 784–788.
Räisänen L, Pohjavirta R, Unkila M & Tuomisto J (1992) A new method for the measurement of
spontaneous motor activity of laboratory animals. Pharmacol Toxicol 70: 230–231.
Ramsey M (1979) Noninvasive automatic determination of mean arterial pressure. Med Biol Eng
Comput 17: 11–18.
Raukola J, Kuusinen N & Paajanen M (2002) Cellular electets-From polymer granules to
electromechanically active films. Proc. 11th International Symposium on Electrets, pp. 195–198.
Reeben V & Epler M (1973) Non-invasive continuous measurement of mean arterial blood
pressure in man. Digest. Int. Conf. on Med Biol Eng, p. 104.
Reeben V & Epler M (1983) Indirect continuous measurement of mean arterial pressure. Advanges
in Cardiovascular Physics 5, Cardiovasc Eng Part II, pp. 90–118.
Rhee S, Yang B-H & Asada, HH (2001) Artifact-resistant power-efficient design of finger-ring
plethysmographic sensors. IEEE Trans Biomed Eng 48(7): 795–805.
82
Rhee S, Yang B-H, Chang K & Asada, HH (1998) The ring sensor: a new ambulatory wearable
sensor for twenty-four hour patient monitoring. Proc. 20th Annual International Conference of
the IEEE-EMBS, Hong Kong, China 20(4): 1906–1909.
Rogueiro-Gómez A & Pallás-Areny R (1998) A new method for automated blood pressure
measurement, Physiol Meas 19: 205–212.
Ruha A (1998) Detection of Heart Beat in Ambulatory Heart Rate Measurements, Methods and
Integrated Circuits. Oulu, Finland: Doctoral thesis, University of Oulu, Department of Electrical
Engineering: 36–42.
Safar ME, London GM, Asmar R & Frohlich ED (1998) Recent advances on large arteries in
hypertension. Hypertension 32: 156–161.
Šantić A & Lacrović I (1999) Simultaneous application of multiple oscillometric methods for blood
pressure measurement in finger. Proc. First Joint BMES/EMBS Conference Serving Humanity,
Advancing Technology, Atlanta, GA, USA: 231.
Šantić A & Šaban M (1997) Two methods for determination of diastolic pressures in fingers. 1995
IEEE-EMBC and CMBEC, Theme 1, Cardiovascular System, Montreal, Que, Canada, 1: 149–
150.
Šantić A, Štritof T & Bilas V (1998) Plethysmography measurements using short current pulses
with low-duty cycle. Proc. 20th Annual International Conference of the IEEE, 20(4): 1889–1892.
Sasayama S, Imachi Y, Yagi T, Imachi K, Ono T & Man-i M (1994) Development of
optoelectronic monitoring system for ear arterial pressure waveforms. Proc. SPIE-The
International Society for Optical Engineering 2084: 354–361.
Sato T, Nishinaga M, Kawamoto A, Ozawa T & Takatsuji H (1993) Accuracy of the continuous
blood pressure monitor based on arterial tonometry. Hypertension 21: 866–874.
Savolainen A & Kirjavainen K (1989) Electromechanical film. Part 1. Design and characteristics. J
Macromolecular Science-Chemistry, 26: 583–591.
Savolainen A (1990) Biaxially oriented polypropylene blown films. I Morphological analysis of
orientation in the machine direction. Polymer Engineering and Science 30: 1258–1264.
Schiller Z & Pasch T (1991) Servo-plethysmo-manometry for continuous noninvasive blood
pressure monitoring. Anaesthesist 40(2): 105–109.
Schmidt TF, Wittenhaus J, Steinmetz TF, Piccolo P & Lüpsen H (1992) Twenty-four-hour
ambulatory noninvasive continuous finger blood pressure measurement with PORTAPRES: a
new tool in cardiovascular research. J Cardiovasc Pharmacol 19 Suppl 6: S117–145.
Sebald DJ, Bahr DE & Kahn AR (2002) Narrowband auscultatory blood pressure measurement.
IEEE Trans Biomed Eng 49(9): 1038–1044.
Segers P, Carlier S, Armstrong G, Verdonck P & Thomas J (1998) In vivo application of the
generalised radial-aorta and carotid-aorta pressure transfer function. Proc. Computers in
Cardiology, Cleveland, OH, USA 1: 669–672.
Shaltis P, Reisner A & Asada H (2005) Calibration of the Photoplethysmogram to Arterial Blood
Pressure: Capabilities and Limitations for Continuous Pressure Monitoring. IEEE-EMBS, Proc.
27th Annual International Conference, Shanghai, China: 3970–3973.
Shaltis P, Wood L, Reisner A & Asada H (2005) Novel Design for a Wearable, Rapidly
Deployable, Wireless Noninvasive Triage Sensor. IEEE-EMBS, Proc. 27th Annual International
Conference, Shanghai, China: 3567–3570.
Siivola J, Leinonen K & Räisänen L (1993) ETMF-polymer transducer as a detector of respiration
in humans. Med Biol Eng Comput 31: 634–635.
Silke B & McAuley D (1998) Accuracy and precision of blood pressure determination with the
Finapres: an overview using re-sampling statistics. J Hum Hypertens 12: 403–409.
Sorvoja HSS & Myllylä R (2003) Accuracy of the electronic palpation method. Proc. 7th
Conference of the European Society for Engineering in Medicine (ESEM), Halle, 1: 196.
83
Stewart WR, Ramsey MW & Jones CJH (1994) Non-invasive measurement of pulse wave velocity
using transputer-based analysis of Doppler flow audio signals. Physiol Meas 15: 299–308.
Streicher, Muller R, Peremans H & Lerch R (2003) Broadband ultrasonic transducer for an
artificial bat head. Proc. IEEE International Ultrasonic Symposium, 2: 1364–1367.
Teng XF & Zhang YT (2003) Continuous and noninvasive estimation of arterial blood pressure
using a photoplethysmographic approach. IEEE-EMBS, Proc. 25th Annual International
Conference, Cancun, Mexico, 4: 3153–3156.
Terry S, Eckerle J, Kornbluh R, Low T & Ablow C (1990) Silicon pressure transducer arrays for
blood-pressure measurement. Sensors and Actuators A21–A23: 1070–1079.
Triedman JK & Saul JP (1994) Comparison of intra-arterial with continuous noninvasive blood
pressure measurement in postoperative pediatric patients. J Clin Monit 10(1): 11–20.
Ursino M & Cristalli C (1994) Mathematical analysis of the oscillometric technique for indirect
blood pressure evaluation. Engineering Advances: New Opportunities for Biomedical
Engineers, Proc. 16th Annual International Conference of the IEEE-EMBS 2: 1286–1287.
Ursino M & Cristalli C (1996) A mathematical study of some biomechanical factors affecting the
oscillometric blood pressure measurement. IEEE Trans Biomed Eng 43(8): 761–778.
van Popele NM, Bos WJW, de Beer NAM, van der Guip DAM, Hofman A, Groppee DE &
Wittemann JCM (2000) Arterial stiffness as underlying mechanism of disagreement between an
oscillometric blood pressure monitor and a sphygmomanometer. Hypertension 36(4): 484–488.
van Popele NM, Grobbee DE, Bots ML, Asmar R, Topouchian J, Reneman RS, Hoeks APG, van
der Kuip DAM, Hofman A & Witteman JCM (2001) Association between arterial stiffness and
atherosclerosis: the Rotterdam study. Stroke 32: 454–460.
Vieri-Gashi E, Kärjä-Koskenkari P, Nissilä S, Sorvisto M, Sorvoja H & Myllylä R (2000) Blood
pressure variation measured by the electronic palpation method and compared to intra-arterial
variation. Proc. Işik 2000 Workshop on Biomedical Information Engineering, Istanbul, Turkey:
101–103.
Vieri-Gashi E, Sorvoja H, Myllylä R, Nissilä S, Sorvisto M, & Kärjä-Koskenkari P (2001) The
effect of the venous pressure to the blood pressure signals measured by the palpation method.
Proc. 23rd Annual International Conference of the IEEE-EMBS, Istanbul, Turkey 4: 3259–3261.
Wang J-J, Lin C-T, Liu S-H & Wen Z-C (2002) Model-based synthetic fuzzy logic controller for
indirect blood pressure measurement. IEEE Trans Systems, Man and Cybernetics, Part B 32(3):
306–315.
Weaver, C S, Eckerle J S, Newgard P M, Warnke C T, Angell J B, Terry SC & Robinson J (1978)
A study of noninvasive blood pressure measurement technique. Noninvasive Cardiovascular
Measurements. Soc Photo-optical Instr Eng 167: 189.
Wegener M & Gerhard-Multhaupt R (2003) Electric poling and electromechanical characterization
of 0.1-mm-thick sensor films and 0.2-mm-thick cable layers from piezoelectric poly(vinylidene
fluoride-trifluoroethylene). IEEE Trans Ultrason Ferroelect Freq Contr 50(7): 921–931.
Wegener M, Paajanen M, Wirges W & Gerhard-Multhaupt R (2002) Corona-induced partial
discharges, internal charge separation and electromechanical transducer properties in cellular
polymer films. Proc. 11th International Symposium on Electrets: 36–45.
Wegener M, Wirges W, Gerhard-Multhaupt R, Paajanen M, Minkkinen H & Raukola J (2003)
Enhancing the cellular structure and the electromechanical response of the ferroelectrets –gas
diffusion expansion of voided polypropylene films. Proc. Annual Report Conference on
Electrical Insulation and Dielectric Phenomena: 36–39.
Wesseling KH (1996) Finger arterial pressure measurement with Finapres. Z Kardiol 85 Suppl 3:
38–44.
84
Williams B, Poulter NR, Brown MJ, Davis M, McInnes GT, Potter JF, Sever PS, Thom SM, BHS
guidelines working party, for the British Hypertension Society (2004) British Hypertension
Society guidelines for hypertension management 2004 (BHS-IV): summary. BMJ 328: 634–
640.
Wood L & Asada H (2005) Active Motion Artifact Reduction for Wearable Sensors Using
Laguerre Expansion and Signal Separation. IEEE-EMBS, Proc. 27th Annual International
Conference, Shanghai, China 1: 3571–3574.
Yang B-H; Rhee S & Asada HH (1998) A twenty-four hour tele-nursing system using a ring
sensor. Robotics and Automation, Proc. IEEE International Conference 1: 387–392.
Yelderman M & Ream A K (1979) Indirect measurement of blood pressure in the anesthetized
patient. Anesthesiol 50: 253–256.
Zheng DC & Zhang YT (2003) A ring-type device for the noninvasive measurement of arterial
blood pressure. IEEE-EMBS, Proc. 25th Annual International Conference, Cancun, Mexico 4:
3184–3187.
Zorn E A, Wilson M B, Angel J J, Zanella J & Alpert B S (1997) Validation of an automated
arterial tonometry monitor using Association for the Advancement of Medical Instrumentation
standards. Blood Press Monitor 2(4): 185–188.
Elvan-Taspinar A, UiterKamp LAMS, Sikkema JM, Bots ML, Koomans HA, Bruinse HW, Franx
A. Validation and use of the Finometer for blood pressure measurement in normal, hypertensive
and pre-eclamptic pregnancy J Hypertension 21: S2053-2060.
Original papers
I
Ruha, A, Miettinen J, Sorvoja H & Nissilä S (1996) Heart rate measurement based
on detection of the radial artery pressure pulse. Proc. BIOSIGNAL’96, Brno, Czech
Republic, 1: 198–200.
II
Sorvoja H, Kokko V-M, Myllylä R & Miettinen J (2005) Use of EMFi as a blood
pressure pulse transducer, IEEE Transactions on Instrumentation and Measurement
54(6): 2505–2512.
III
Nissilä S, Sorvisto M, Sorvoja H, Vieri-Gashi E & Myllylä R (1998) Noninvasive
blood pressure measurement based on the electronic palpation method. Proc. 20th
Annual International Conference of the IEEE Engineering in Medicine and Biology
Society, Chicago, USA, 20(4): 1723–1726.
IV
Sorvoja H, Myllylä R, Nissilä S, Kärjä-Koskenkari P, Koskenkari J, Lilja M &
Kesäniemi YA (2001) A method to determine diastolic blood pressure based on
pressure pulse propagation in the electronic palpation method, Proc. 23rd Annual
International Conference of the IEEE Engineering in Medicine and Biology Society,
Istanbul, Turkey, 4: 3255–3258.
V
Sorvoja H, Myllylä R, Nissilä S, Kärjä-Koskenkari P, Koskenkari J, Lilja M &
Kesäniemi YA (2001) Systolic blood pressure accuracy enhancement in the
electronic palpation method using pulse waveform, Proc. 23rd Annual International
Conference of the IEEE Engineering in Medicine and Biology Society, Istanbul,
Turkey, 1: 222–225.
VI
Sorvoja H, Hast J, Myllylä R, Kärjä-Koskenkari P, Nissilä S & Sorvisto M (2003)
Blood pressure measurement method using pulse-transit-time. Molecular and
Quantum Acoustics 24: 169–181.
VII Sorvoja H & Myllylä R (2004) Accuracy of the electronic palpation blood pressure
measurement method versus the intra-arterial method. Technology and Health Care,
International Journal of Health Care Engineering 12(2): 145–146.
VIII Sorvoja H, Myllylä R, Kärjä-Koskenkari P, Koskenkari J, Lilja M & Kesäniemi YA
(2005) Accuracy comparison of oscillometric and electronic palpation blood
pressure measuring methods using intra-arterial method as a reference. Molecular
and Quantum Acoustics 26: 235–260.
86
IX
Hast J, Myllylä R, Sorvoja H & Miettinen J (2002) Arterial pulse wave shape
measurement using self-mixing effect in a diode laser, Quantum Electronics 32(11):
975–980.
Original publications are not included in the electronic version of the dissertation.
C259etukansi.fm Page 2 Tuesday, November 28, 2006 9:55 AM
ACTA UNIVERSITATIS OULUENSIS
SERIES C TECHNICA
242.
Lumijärvi, Jouko (2006) Optimization of critical flow velocity in cantilevered fluidconveying pipes, with a subsequent non-linear analysis
243.
Stoor, Tuomas (2006) Air in pulp and papermaking processes
244.
György, Zsuzsanna (2006) Glycoside production by in vitro Rhodiola rosea
cultures
245.
Özer-Kemppainen, Özlem (2006) Alternative housing environments for the
elderly in the information society. The Finnish experience
246.
Laurinen, Perttu (2006) A top-down approach for creating and implementing data
mining solutions
247.
Jortama, Timo (2006) A self-assessment based method for post-completion audits
in paper production line investment projects
248.
Remes, Janne (2006) The development of laser chemical vapor deposition and
focused ion beam methods for prototype integrated circuit modification
249.
Kinnunen, Matti (2006) Comparison of optical coherence tomography, the pulsed
photoacoustic technique, and the time-of-flight technique in glucose
measurements in vitro
250.
Iskanius, Päivi (2006) An agile supply chain for a project-oriented steel product
network
251.
Rantanen, Rami (2006) Modelling and control of cooking degree in conventional
and modified continuous pulping processes
252.
Koskiaho, Jari (2006) Retention performance and hydraulic design of constructed
wetlands treating runoff waters from arable land
253.
Koskinen, Miika (2006) Automatic assessment of functional suppression of the
central nervous system due to propofol anesthetic infusion. From EEG
phenomena to a quantitative index
254.
Heino, Jyrki (2006) Harjavallan Suurteollisuuspuisto teollisen ekosysteemin
esimerkkinä kehitettäessä hiiliteräksen ympäristömyönteisyyttä
255.
Gebus, Sébastien (2006) Knowledge-based decision support systems for
production optimization and quality improvement in the electronics industry
256.
Alarousu, Erkki (2006) Low coherence interferometry and optical coherence
tomography in paper measurements
257.
Leppäkoski, Kimmo (2006) Utilisation of non-linear modelling methods in fluegas oxygen-content control
Book orders:
OULU UNIVERSITY PRESS
P.O. Box 8200, FI-90014
University of Oulu, Finland
Distributed by
OULU UNIVERSITY LIBRARY
P.O. Box 7500, FI-90014
University of Oulu, Finland
C259etukansi.fm Page 1 Tuesday, November 28, 2006 9:55 AM
C 259
OULU 2006
UNIVERSITY OF OULU P.O. Box 7500 FI-90014 UNIVERSITY OF OULU FINLAND
U N I V E R S I TAT I S
S E R I E S
SCIENTIAE RERUM NATURALIUM
Professor Mikko Siponen
HUMANIORA
Professor Harri Mantila
TECHNICA
Professor Juha Kostamovaara
ACTA
U N I V E R S I T AT I S O U L U E N S I S
Hannu Sorvoja
E D I T O R S
Hannu Sorvoja
A
B
C
D
E
F
G
O U L U E N S I S
ACTA
A C TA
C 259
NONINVASIVE BLOOD
PRESSURE PULSE DETECTION
AND BLOOD PRESSURE
DETERMINATION
MEDICA
Professor Olli Vuolteenaho
SCIENTIAE RERUM SOCIALIUM
Senior Assistant Timo Latomaa
SCRIPTA ACADEMICA
Communications Officer Elna Stjerna
OECONOMICA
Senior Lecturer Seppo Eriksson
EDITOR IN CHIEF
Professor Olli Vuolteenaho
EDITORIAL SECRETARY
Publications Editor Kirsti Nurkkala
ISBN 951-42-8271-X (Paperback)
ISBN 951-42-8272-8 (PDF)
ISSN 0355-3213 (Print)
ISSN 1796-2226 (Online)
FACULTY OF TECHNOLOGY,
DEPARTMENT OF ELECTRICAL AND INFORMATION ENGINEERING,
INFOTECH OULU,
UNIVERSITY OF OULU
C
TECHNICA