1
AMERICAN UNIVERSITY OF SHARJAH
SCHOOL OF ENGINEERING
COMPUTER SCIENCE & ENGINEERING
Fuzzy Logic based
Patients’ Monitoring System
Presented by :
Student Name : Jumanah A. Al-Dmour
Supervised by :
Advisor
: Prof. Abdulrahman Al-Ali
Co-advisors : Prof. Assim Sagahyroon
: Prof. Salah Abusnana
Fall 2012 / 2013
Outlines
2
Introduction
General Problem
Literature Review
Overall System Design
Software Architecture
Research Implementation and Results
Conclusion
Introduction
3
 The number of older persons has tripled over the last 50 years; it will more
than triple again over the next 50 years.
 The older population is growing faster than the total population in
practically all regions of the world―and the difference in growth rates is
increasing.
Average annual growth rate of total population and population aged 60 or over [1]
General Problem
4
 In general, life expectancy is increasing, this will lead to a tremendous increase in
aging population.
 A much different set of expectations of quality of life and medical care.
 Many academic institutions and industrial organizations are engaged in
healthcare research.
 Philips, Intel, GE Medical, IBM, Medtronic, and Carnegie Mellon.
Literature Review
Patients’ Monitoring Systems
5
 Patient monitoring system is a system that consists of various devices that
are used to monitor and supervise patients and alerts if the patient gets into
a critical state such as a heart monitor.
 Why Wireless?
 cost effectiveness.
 Provide a better Quality of Life.
 Patients mobility.
 User or patient's ability to view his/her medical data trends anywhere and
anytime with minimum additional hardware requirements
Literature Review
Existing Wireless Technologies
6
Zigbee-Based Solution
WiFi-Based Solution
Literature Review
Fuzzy Logic Based Systems
7
A closed loop feedback system
Continuously monitors the patient’s
blood glucose level and adjusts the
infusion of insulin to an optimal
rate
Blood glucose monitoring
using Fuzzy logic
Overall System Design
8
 In this research, our goal is to:
 Design, build, and test a wireless data acquisition unit DAQ to collect patient
vital signs while they are on the move.
 Develop a DAQ API and database to profile patients and save their medical
records and health status.
 Develop a Fuzzy Logic algorithm based on the MEWS system to online analyze
the patients’ vital signs and issue warnings and send alarm messages to caretaker
in-case of any abnormality.
 Implement and test the proposed system.
Overall System Design
9
 Functional Requirements
 Arranging readers in a specific arrangement and localizing their
positions to a fixed dataset,
 Performing registration tasks,
 Collecting patients’ vital signs
 Storing data
 Alerting staff
 Non-Functional Requirements
Accessibility
Scalability
Security
Software usability
Safety
Operational requirements
Overall System Design
Overall System Architecture
10
Based on the requirements, the proposed wireless monitoring
system consists of five major building modules:
Overall System Design
The Mobile Data Acquisition Unit
11
 The mobile data acquisition unit module consists of
the RFID based vital signs sensors.
 The system consists of the following RFID based
vital signs sensors:
 Blood Pressure Sensor
 Pulse Oxi-meter (SPO2)
 Body Temperature Sensor
 Blood Sugar (Glucose) Sensor
Overall System Design
12
Software
Architecture
13
Software
Architecture
14
Software
Architecture
15
Software Architecture
The Fuzzy Logic System
16
 The concept of Fuzzy Logic (FL) was first conceived by Lotfi Zadeh, a
professor at the University of California at Berkley. [1]
 The concept was presented not as a control methodology, but as a way of
processing data by allowing partial set membership rather than crisp set
membership or NonMembership.
 Professor Zadeh reasoned that people do not require
precise, numerical information input, and yet they are
capable of highly adaptive control.
 If feedback controllers could be programmed to accept
noisy, imprecise input, they would be much more
effective and perhaps easier to implement.
[1] Diversity Tech – FPGA & BOARD DESIGN SERVICES
Software Architecture
The Fuzzy Logic System
17
Crisp Set Vs. Fuzzy Set
Crisp logic needs hard decisions. Like in this chart. In this example, anyone lower
than 175 cm considered as short, and behind 175 considered as high. Someone
whose height is 180 is part of TALL group, exactly like someone whose height is
190.
Fuzzy Logic deals with “membership
in group” functions. In this example,
someone whose height is 180, is a
member in both groups. Since his
membership in group of TALL is 0.5
while in group of SHORT only 0.1, it
may be seen that he is much more
TALL than SHORT.
Software Architecture
The Fuzzy Logic System
18
 Typical Fuzzy Logic control implementation involving 3 steps:
Sensors
Inputs
System
Status
Fuzzification
Rule Proc. #1
Rule Proc. #2
Rule Proc. #3
Rule Proc. #4
…..
Rule Proc. #n
Control
Response
Action
Outputs
De-Fuzzification
 Fuzzification: converting the crisp inputs to membership functions which comply
to intuitive perception of system status.
 Rules Processing: calculating the response from system status inputs according to
the pre-defined rules matrix (control algorithm implementation).
 De-Fuzzification: converting the Rules Processing results to crisp output/s to feed
into the control devices.
Software Architecture
The Modified Early Warning
Score System Mews
19
The Modified Early Warning Score (MEWS) is a tool for bedside evaluation of patients
and is based on five physiological parameters: systolic blood pressure, pulse rate,
respiratory rate, temperature, and AVPU score (A for 'alert', V for 'responsive to verbal
stimulation', P for 'responsive to painful stimulation', U for 'unresponsive').
Software Architecture
The Modified Early Warning
Score System Mews
20
 In this work, additional MEWS parameters were used in order to calculate the
MEWS score.
 The parameters used are:
 Systolic Blood Pressure (SBP),
 Heart Rate (HR),
 Oxygen saturation (SPO2),
 Body Temperature (TEMP),
 and Blood Sugar (BS)
Modified Early Warning Score MEWS
Risk Band
Low
Low
Low
Normal
High
High
High
Vital Sign
+3
+2
+1
+0
+1
+2
+3
Systolic Blood Pressure
SBP<75
70<SBP<85
80<SBP<100
95<SBP<199
HR<50
45<HR<60
53<HR<100
83<SPO2<90
87<SPO2<95
SPO2>93
Heart Rate
SPO2
SPO2<85
Temperature
Blood Sugar Diabetes
T<36.5
BS<66
63<BS<72
SBP>185
95<HR<110
36<T<38.5
-
70<BS<110
105<HR<130
HR>125
T>38
-
106<BS<150
BS>140
Software Architecture
Structure Of The Fuzzy Logic
System
21
Software Architecture
Fuzzy Expert System Designing
22
 The typical steps followed in designing expert systems include:
 The determination of the input and output variables,
 The selection of suitable membership functions,
 and The creation of the fuzzy rules database.
Software Architecture
Fuzzy Expert System Designing
Input Variables
23
 An actual input (Blood Sugar) may have one or more values that are
 Blood Sugar: Blood sugar input is a very important factor. This input
associated with more than one membership function. For example:
field
has five
(Low-3,
Low-2,Normal/1,
Normal-0,High2/0,
High-2, and
High-3).
 Input:
80 fuzzy
– hassets
(Low3/0,
Low2/0,
High3/0)
 Input: 143 – has (Low3/0, Low2/0, Normal/1, High2/0.3, High3/0.7)
1
Input Field
0.7
Range
Fuzzy Sets
<66
Low-3
63 - 72
Low-2
70 - 110
Normal-0
106 - 150
High-2
>140
High-3
Blood Sugar
BS
0.3
63 66 70 72
80
106 110
140 143
150
Software Architecture
Fuzzy Expert System Designing
Input Variables
24
 Blood Pressure: In this field, the systolic blood pressure SBP was used.
This input variable is divided into 5 fuzzy sets. These sets are (Low-3,
Low-2, Low-1, Normal-0, and High-2)
Input Field
Range
Fuzzy Sets
<75
Low-3
Systolic Blood
70 – 85
Low-2
Pressure
80 – 100
Low-1
SBP
95 – 199
Normal-0
>185
High-2
Software Architecture
Fuzzy Expert System Designing
Input Variables
25
 Heart Rate: based on the MEWS scoring system and per the expert
advice, we use for this input field six fuzzy sets (Low-2, Low-1,
Normal-0, High-1, High-2, and High-3).
Input Field
Range
Fuzzy Sets
<50
Low-2
45 - 60
Low-1
Heart Rate
53 - 100
Normal-0
HR
95 – 110
High-1
105 - 130
High-2
>125
High-3
Software Architecture
Fuzzy Expert System Designing
Input Variables
26
 SPO2: The value of this input field is the oxygen saturation in the
patient’s blood. In this field, we have 4 linguist variables (fuzzy sets)
(Low-3, Low-2, Low-1, and Normal).
Input Field
Range
Fuzzy Sets
<85
Low-3
83 - 90
Low-2
87 - 95
Low-1
>93
Normal-0
SPO2
Software Architecture
Fuzzy Expert System Designing
Input Variables
27
 Temperature: In this field, we have 3 fuzzy sets (Low-2, Normal-0, and
High-2).
Input Field
Range
Fuzzy Sets
<36.5
Low-2
36 – 38.5
Normal-0
>38
High-2
Temperature
TEMP
Software Architecture
Fuzzy Expert System Designing
Output Variables
28
 There is one output variable “Risk Group”,
Output Field
Range
Fuzzy Sets
0<RG<0.5
NRM
0.5<RG<1.5
LRG1
1.5<RG<2.5
LRG2
2.5<RG<3.5
LRG3
 In this system, we have 15 fuzzy sets for the
3.5<RG<4.5
LRG4
4.5<RG<5.5
HRG5
output variable risk group (NRM, LRG1,
5.5<RG<6.5
HRG6
6.5<RG<7.5
HRG7
7.5<RG<8.5
HRG8
8.5<RG<9.5
HRG9
9.5<RG<10.5
HRG10
10.5<RG<11.5
HRG11
11.5<RG<12.5
HRG12
12.5<RG<13.5
HRG13
13.5<RG<14
HRG14
which refers to the degree of risk in a patient’s
case. It ranges from 0 to 15.
Risk Group
NRM
LRG2, LRG3, LRG4, HRG5, HRG6, HRG7,
HRG8, HRG9, HRG10, HRG11, HRG12,
LRG
HRG13, and HRG14).
HRG
Software Architecture
Fuzzy Expert System Designing
The Fuzzy Rule Base
29
 The rules were designed based on the MEWS scoring system. And the
results with the 1800 rules tend to be similar to the MEWS scoring
system.
 The numbers of rules were obtained using the following formula:
N = p1 ×p2 ×………×pn
Where
[13]
N is the total number of possible rules for a fuzzy system
PN is the number of linguistic terms for the input linguistic variable N.
Software Architecture
Fuzzy Expert System Designing
The Fuzzy Rule Base
30
Modified Early Warning Score MEWS
Risk Band
Low
Low
Low
Normal
High
High
High
Vital Sign
+3
+2
+1
+0
+1
+2
+3
Systolic Blood Pressure
SBP<75
70<SBP<85
80<SBP<100
95<SBP<199
HR<50
45<HR<60
53<HR<100
83<SPO2<90
87<SPO2<95
SPO2>93
Heart Rate
SPO2
SBP<85
Temperature
Blood Sugar Diabetes
T<36.5
SBP<66
63<BS<72
SBP>185
95<HR<110
36<T<38.5
-
70<BS<110
105<HR<130
HR>125
T>38
-
106<BS<150
>140
 The numbers of rules were obtained using the following formula:
N
= p𝑆𝐵𝑃 × p𝐻𝑅 × p𝑆𝑃𝑂2 × p 𝑇𝐸𝑀𝑃 × p𝐵𝑆
= 5 × 6 × 4 × 3 × 5 = 1800
Software Architecture
Fuzzy Expert System Designing
The Fuzzy Rule Base
31
Software
Architecture
32
Software
Architecture
33
Implementation, Results
and Discussion
34
 The proposed hardware validation was carried in Rashid Center for Diabetes
and Research (RCDR) at Khalifa Hospital in Ajman under the supervision and
approval of Professor Salah Abusnana, the medical director of the center.
 A data collection design was used to gather data on participants of different
ages. We measured the participant's blood pressure (BP) level, heart rate (HR),
Oxygen saturation level in blood (SPO2), temperature (TEMP) level, and blood
sugar (BS) level.
Implementation, Results
and Discussion
Hardware Sensors Validation
35
 Blood Pressure (BP) Measurements
Differences between RCDR and RFID BP devices (mmHg)
SBP
DBP
BP Device
Mean ± s.d.
Difference ±
s.d.
≤5 mmHg
≤10
mmHg
≤15
mmHg
Grade
RCDR BP
Device
RFID BP
Device
RCDR BP
Device
RFID BP
Device
146.9565 ±
25.1332
145.0870 ±
24.7311
1.8696 ±
7.4852
65%
86.9%
95.65%
B
-4.5217 ±
5.2559
66%
91.3%
95.65%
B
77.087 ± 10.5310
81.6087 ±
11.9611
Implementation, Results
and Discussion
Hardware Sensors Validation
36
 Blood Pressure (BP) Measurements
Implementation, Results
and Discussion
Hardware Sensors Validation
37
 Blood Sugar (BS) Measurements
Implementation, Results
and Discussion
Hardware Sensors Validation
38
 SPO2 (Oxygen Saturation Level in Blood) Measurements
SPO2 sensors
Range %
Mean %
SD %
RCDR SPO2
sensor
94 - 100
97.0769
1.3365
RFID SPO2
sensor
94 - 99
97.1154
1.4120
Bias %
(Difference)
Precision %
(SD)
Root-mean-square
deviation % (RMSD)
-0.0385
1.4827
1.454436
Implementation, Results
and Discussion
Hardware Sensors Validation
39
 Heart Rate (Pulse) Measurements
Measurement
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
RCDR HR sensor
(bpm)
62
70
92
58
62
82
121
92
76
87
84
92
84
96
89
82
63
70
95
75
65
79
87
92
71
88
RFID BP HR sensor
(bpm)
64
71
88
61
67
80
118
95
79
86
79
93
86
99
92
85
65
71
91
69
64
78
90
91
71
87
RFID SPO2 HR sensor
(bpm)
69
66
90
59
63
80
118
95
73
82
80
93
91
99
93
83
60
71
87
80
65
80
90
91
71
86
Implementation, Results
and Discussion
Hardware Sensors Validation
40
 Heart Rate (Pulse) Measurements
Regression Statistics
Correlation Coefficient
Coefficient of Determination
Difference Standard Deviation (SD)
2xSD
RFID BP HR device
RFID SPO2 HR device
0.9787
0.9579
2.9163
5.8326
0.9658
0.9328
3.6494
7.2988
Implementation, Results
and Discussion
Hardware Sensors Validation
41
 Temperature Measurements
The reading of the wristband temperature differs from the usual body
temperature. Short experiments were conducted to examine and determine
the correction factor of the wristband temperature sensor.
Time
15 minutes
30 minutes
45 minutes
60 minutes
75 minutes
90 minutes
105 minutes
120 minutes
Thermometer value
(Oral Temperature)
RFID Temperature sensor value
(Skin Temperature)
Difference
Correction Factor
°C
°C
°C
36.7
36.8
36.8
36.8
36.6
36.7
36.7
36.7
30.8
31.6
32.24
32.26
32.05
31.9
32
32
5.9
5.2
4.56
4.54
4.55
4.8
4.7
4.7
Results and Discussion
Implementations, Testing, And Evaluation
Of The Fuzzy Logic Engine
42
The evaluation of the fuzzy expert system is the result of the
comparison between the risk groups status suggested by the fuzzy
expert system (decision support system), and the risk groups status
indicated by the MEWS scoring system.
Results and
Discussion
43
IMPLEMENTATIONS, TESTING,
AND EVALUATION OF THE
FUZZY LOGIC ENGINE
Results and Discussion
Implementations, Testing, And Evaluation
Of The Fuzzy Logic Engine
44
Results and Discussion
NRM
45
Normal Group (NRM):
 The system was tested using 5 cases.
 The first 4 cases are similar in having normal vital signs as inputs (the
five inputs are normal), while the 5th case is with 4 normal vital signs
and 1 fuzzy vital sign..
Normal
Range
Results and Discussion
LRG
46
 Low Risk Group (LRG):
 In this category, the system was tested using 19 cases.
 The first five cases are under the LRG 2 category, and the 14 following
cases are under the LRG 3 Category.
LRG
Range
LRG
Range
Results and Discussion
HRG
47
 High Risk Group (HRG):
 In this category, the system was tested using 10 cases.
HRG
Range
Conclusions
48
To conclude:
 A mobile patient monitoring system was developed to monitor patients
while they are on the move.
 A database, a GUI, and a web application were developed to profile
patients and archive their health status.
 A Fuzzy Logic base system was developed to get an alternative early
warning system compared with the existing MEWS.
 The proposed system was tested under the supervision of RCDR:
 The proposed system RFID sensors were tested and compared with
the RCDR devices. The error percentages were within the
international acceptable error ranges.
 26 real time patients from RCDR were tested using the system
 Our system results were classified within the same category of the
MEWS. Though, the Fuzzy logic base early warning system
outperformed the MEWS system, since it came up with more precise
results.
Future Works
49
Possible future research directions include:
 Increasing the sample size from 26 patients to a larger set to
enhance the testing and accuracy of the system. A larger
population of patients will provide many alternate illness
scenarios and hence a better coverage of all possibilities.
 To be able to develop the Fuzzy Logic System to know the
abnormal factors (Ex. Temperature).
50
Thank You
Any Questions?
Results and
Discussion
51
Both patient 4 and patient 5 are patients with 3 normal inputs [SBP, HR, and
SPO2] and 2 abnormal inputs [temperature and BS]; however, they have a
different fuzzy logic score.
RCDR
Patient 4
Patient 5
Temp
36.1
38.2
SBP
Input
Variable
LOW3
LOW2
LOW1
NORMAL0
HIGH2
Membership
value
0
0
0
1
0
Membership
value
0
0
0
1
0
HR
88
63
HR
Input
Variable
LOW2
LOW1
NORMAL0
HIGH1
HIGH2
HIGH3
SBP
Input
Variable
LOW3
LOW2
LOW1
NORMAL0
HIGH2
SBP
129
117
Membership
value
0
0
1
0
0
0
HR
Input
Variable
LOW2
LOW1
NORMAL0
HIGH1
HIGH2
HIGH3
Membership
value
0
0
1
0
0
0
Using MEWS
SPO2
97
98
BS
134.28
115
Status
LRG
LRG
SPO2
Input
Variable
LOW3
LOW2
LOW1
NORMAL0
Membership
value
0
0
0
1
SPO2
Input
Variable
LOW3
LOW2
LOW1
NORMAL0
Membership
value
0
0
0
1
Score %
2
2
Using Fuzzy With
Overlap
Status
Score %
LRG
3.948
LRG
3.378
TEMP
Input
Variable
LOW2
NORMAL0
HIGH2
Membership
value
0.8
0.2
0
BS
Input
Variable
LOW3
LOW2
NORMAL0
HIGH2
HIGH3
TEMP
Input
Variable
LOW2
NORMAL0
HIGH2
Membership
value
0
0.6
0.4
Membership
value
0
0
0
1
0
BS
Input
Variable
LOW3
LOW2
NORMAL0
HIGH2
HIGH3
Membership
value
0
0
0
1
0
Results and
Discussion
52
The defuzzification of the data into a crisp output is accomplished by combining the
results of the inference process and then computing the "fuzzy centroid" of the area.
𝑅𝑖𝑠𝑘𝐺𝑟𝑜𝑢𝑝 =
( 0.8×4.5 + 0.2×2.5 )
(0.8+0.2)
= 4.1
----- Equation 1
𝑅𝑖𝑠𝑘𝐺𝑟𝑜𝑢𝑝 =
( 0.6×2.5 + 0.4×4.5 )
(0.6+0.4)
= 3.3
----- Equation 2
Rule #
Operators
Strength
Centroid
Rule 1249
If (SBP is Normal0) and (HR is Normal0) and (SPO2 is
Normal0) and (TEMP is Low2) and (BS is High2) then
(RiskGroup is LRG4)
1&1&1&0.8&1 = 0.8
LRG4  0.8
Centroid = 4.5
Rule 1254
If (SBP is Normal0) and (HR is Normal0) and (SPO2 is
Normal0) and (TEMP is Normal0) and (BS is High2) then
(RiskGroup is LRG2)
1&1&1&0.2&1 = 0.2
LRG2 0.2
Centroid = 2.5
Rule #
Operators
Strength
Centroid
Rule 1254
If (SBP is Normal0) and (HR is Normal0) and (SPO2 is
Normal0) and (TEMP is Normal0) and (BS is High2) then
(RiskGroup is LRG2)) (1)
1&1&1&0.6&1 = 0.6
LRG2  0.6
Centroid = 2.5
Rule 1259
If (SBP is Normal0) and (HR is Normal0) and (SPO2 is
Normal0) and (TEMP is High2) and (BS is High2) then
(RiskGroup is LRG4)
1&1&1&0.4&1 = 0.4
LRG4 0.4
Centroid = 4.5
Results and
Discussion
53
The simulated and calculated results are shown in the Table 29. The difference
between the results is due to the involved software’s delay.
Results
Patient 4
Patient 5
MATLAB Simulation
3.948
3.378
Calculated Values
4.1
3.3
Difference
0.512
0.078
Error %
1.75
0.34