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