Volunteer Study for Sensor Calibration
To calibrate the new SpO2 sensor family it was necessary to adjust the relationship between the ratio measurements and the SpO2
values using data based on real blood samples from volunteers.
Fig. 1 shows the measurement environment for the calibration study. The basic instrument is a special HP Component Monitoring
System (CMS) with 16 SpO2 channels. Sixteen sensors at different application sites could be used simultaneously. To get SpO2 values
over the entire specification range of 70%SpO2100%, the volunteers got air-nitrogen mixtures with lowered oxygen levels—less
than 21%.
HP Component
Monitoring System
16 SpO2 Channels
10%
O2
Laptop
Computer
Regression
Analysis
Computer
Data
Files
Arterial Blood
Sample
21%
Nitrogen
Oxygen
Valve
OSM3
Oximeter
Fig. 1. Sensor calibration using volunteers and SpO2 data acquisition by a special HP Component Monitoring
System (CMS) with a maximum of 16 channels. Different SpO2 values are achieved by supplying different mixtures of oxygen and nitrogen. Arterial blood samples are analyzed by a Radiometer OSM3 oximeter. For each
sensor and application site, regression analysis is done, and calibrating tables are derived from the results.
Because of his great experience with such studies we used the method developed by Dr. J.W. Severinghaus of the University of
California in San Francisco. For each volunteer a maximum of 16 sensors were applied at the fingers, earlobes, and nostrils. A catheter
was placed in the left radial artery. Arterial O2 saturation was reduced rapidly by a few breaths of 100% N2. This was followed by a
mixture of air and N2 with about 4% CO2 added while the subject voluntarily hyperventilated to speed the attainment of an alveolar gas
hypoxic plateau and to provide end tidal samples for regression analysis. FiO2 was adjusted to obtain plateaus for 30 to 60 seconds at
different SpO2 levels (Fig. 2). At the end of each plateau a 2-ml arterial blood sample was obtained and analyzed by a Radiometer OSM3
multiwavelength oximeter.
The regression analysis yielded three SpO2-versus-ratio calibration curves: one for the HP M1190A adult sensor, a second for the HP
M1191A adult sensor, the HP M1192A pediatric sensor, and the HP M1193A neonatal sensor, and a third for the HP M1194A ear sensor.
The curves for the HP M1190A and M1191A are different because of their different LED wavelengths, while for the ear sensor the
application site is different— the tissue constitution of the earlobe and nostril seems to be optically very different from the other
application sites. Each calibration curve is the best least squares fit to the data points of a second-order polynomial.
Fig. 3 shows the good correlation with the reference (R2 0.95) in the case of the HP M1191A adult sensor. Fig. 4 shows that the
specified SpO2 accuracy is reached within the range of 70%SpO2100%. Fig. 5 shows that the correlation for the HP M1194A ear
sensor is not as good as for the HP M1191A. The data point distribution is also wider (Fig. 6). This is caused by a much poorer signal
quality at the earlobe than at the finger. In general the perfusion index for the ear is only about a tenth of that for the finger. Therefore, in
normal circumstances the preferred application site is the finger. In some cases, such as centralization (i.e., shock patients), the earlobe
sometimes gives better results.
Subarticle 7a
February 1997 Hewlett-Packard Journal
1
100
20
FiO2
SpO2
15
80
10
60
5
40
0
0
t1
1
t2
2
t3 3 t4
Time t (min)
4
5
6
80
HP M1191A SpO 2 Reading (%)
Blood
Samples
Arterial Blood Oxygen Saturation Measured
Noninvasively with a Pulse Oximeter SpO 2 (%)
Fractional Inspired Oxygen Concentration
in Breathing Air FiO2 (%)
100
60
40
20
Fig. 2. Stepwise desaturation by lowering oxygen levels
leads to quasistable SpO2 levels. This condition gives
blood samples with correct blood gas values. The delay
for the SpO2 values compared to the oxygen values
comes from the circulation time for the arterial blood
from the lungs to the arm. Calibration tables are compiled
by comparing the known SpO2 values with the ratio data
measured by the CMS.
20
40
60
OSM3 SaO2 Reading (%)
80
Fig. 3. Regression analysis for HP M1191A adult sensor SpO2
measurements after calibration. The measurements are
plotted against arterial blood SaO2 measurements from the
OSM3 oximeter. The data (206 points) is from 12 volunteers
with different oxygen saturation levels.
100
HP M1194A SpO 2 Reading (%)
20
Bias
SpO 2
SaO 2 (%)
15
10
5
0
–5
80
60
40
–10
–15
–20
50
60
80
70
OSM3 SaO2 Reading (%)
90
Fig. 4. Bias and standard deviation for the HP M1191A over
the specified range of 70%SpO2 100%.
Subarticle 7a
100
20
20
40
60
OSM3 SaO2 Reading (%)
80
100
Fig. 5. Regression analysis for HP M1194A ear sensor SpO2
measurements after calibration. The measurements are
plotted against arterial blood SaO2 measurements from the
OSM3 oximeter for 12 volunteers.
February 1997 Hewlett-Packard Journal
2
20
SaO 2 (%)
10
SpO 2
15
0
5
Bias
–5
–10
–15
–20
50
60
70
80
OSM3 SaO2 Reading (%)
90
100
Fig. 6. Bias and standard deviation for the HP M1194A. The standard deviation is larger
than for the HP M1191A because of the smaller perfusion values in the ear.
Subarticle 7a
Go to Subarticle 7b
Return to Article 7
Go to Next Article
Go to Journal Home Page
February 1997 Hewlett-Packard Journal
3