DESIGN AND ANALYSIS OF ARTIFACT-RESISTIVE FINGER PHOTOPLETHYSMOGRAPHIC SENSORS FOR VITAL SIGN MONITORING by Sokwoo Rhee B.S., Mechanical Engineering Seoul National University, Korea (1995) M.S., Mechanical Engineering Massachusetts Institute of Technology (1997) Submitted to the Department of Mechanical Engineering in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy at the BARKER Massachusetts Institute of Technology June 2000 MASSACHUSE1fT TC OF TECHNOLOGY ©2000 Massachusetts Institute of Technology All rights reserved LIBRARIES Signature of Author Department of Mechanical Engineering May 8, 2000 Certified by 'Y Harry H. Asada --7 Professor>fMechanical Engineering Thesis Supervisor Certified by Boo-Ho Yang Lntist of Mechanical Engineering rhesis Co-supervisor Accepted by____________ ________ Ain A. Sonin Chairman, Department Committee on Graduate Students DESIGN AND ANALYSIS OF ARTIFACT-RESISTANT FINGER PHOTOPLETHYSMOGRAPHIC SENSORS FOR VITAL SIGN MONITORING by Sokwoo Rhee Submitted to the Department of Mechanical Engineering On May 8, 2000, in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Mechanical Engineering ABSTRACT A miniaturized, telemetric, photoplethysmograph sensor for long-term, continuous monitoring is presented in this thesis. The sensor, called a "ring sensor," is attached to a finger base for monitoring beat-to-beat pulsation, and the data is sent to a host computer via a RF transmitter. Two major design issues are addressed: one is to minimize motion artifact and the other is to minimize the consumption of battery power. An efficient double ring design is developed to lower the influence of external force, acceleration, and ambient light, and to hold the sensor gently and securely on the skin, so that the circulation at the finger may not be obstructed. To better understand the mechanism of motion artifact by external forces, a comprehensive mathematical model describing the finger photoplethysmography was developed and verified by finite element method, numerical simulation and experiments. Total power consumption is analyzed in relation to the characteristics of the individual components, sampling rate, and CPU clock speed. Optimal operating conditions are obtained for minimizing the power budget. A prototype ring sensor is designed and built based on the power budget analysis and the artifact-resistive attachment method. It is verified through experiments that the ring sensor is resistant to interfering forces and acceleration acting on the ring body. It is also shown that the device meets diverse and conflicting requirements, including compactness, motion artifact reduction, minimum loading effects, and low battery power consumption. Benchmarking tests with FDA-approved photoplethysmograph and EKG reveal that the ring sensor is comparable to those devices in detecting beat-to-beat pulsation despite disturbances. The long-term monitoring experiment shows that this device can effectively provide a considerable amount of artifact-free vital sign information in everyday life. Finally, guidelines for designing the ring sensor are proposed based on the analyses and the experiment results. Thesis Committee Members: Professor Harry H. Asada, Chairman Professor Roger D. Kamm Professor Roger G. Mark Dr. Boo-Ho Yang 2 To My Lovely Family... 3 Acknowledgements It has already been five years since I stood in front of the main gate of MIT for the first time. During that time, so many things have happened to me. Some of them were very exciting and delighting, and some of them were sad and discouraging. After all, I am so glad that I could finish my Ph.D. work and write this thesis. Most of all, I would like to express my best and sincere thanks to my thesis advisor, Professor Harry H. Asada, for his constant encouragement and guidance. His profound insight and splendid wide vision gave me a great chance to get into the world of new research directions. His valuable support and advice were the greatest factor that enabled me to write this thesis. I also would like to express deep gratitude to the thesis committee members: Professor Roger D. Kamm, Professor Roger G. Mark, and Dr Boo-Ho Yang. They guided and helped me a lot in my research, and contributed greatly to my thesis. I would like to express thanks to all my lab-mates in d'Arbeloff Laboratory who showed me sincere friendship and care. Also I would like to express deep thanks to the continuous support of my good friends, especially Sangjun Han and Andy S. Kim. They made my life more energetic and enjoyable while I was going through a hard time struggling to make progress in my research. In addition, I would like to say thanks to all my friends at MIT. Finally, I would like to give my best appreciation to my lovely wife, Eunkyoung Um. She has always been a great supporter of my work. I also would like to say thanks to my parents and my sister, who have been watching me with great love. Their love and care have been the main source of energy that has encouraged me throughout my life at MIT. Even after I graduate from MIT, I will not be able to forget this wonderful school, and I think what MIT has taught me these five years will be the major thrust that will guide me through the rest of my life. 4 Contents 1. IN TR O D U CT IO N ................................................................................................................... 9 1.1 BACKGROUND AND OBJECTIVES ..................................................................................................... 9 1.2 PRIOR W ORK IN THE FIELD................................................................................................................ 10 1.3 OUTLINE OF THESIS ........................................................................................................................... 11 2. TH E R IN G SEN SO R ............................................................................................................ 14 2.1 BASIC DESCRIPTION OF THE RING SENSOR ....................................................................................... 14 2 .2 IS SU E S ................................................................................................................................................ 15 3. A R TIFA C T -R E SISTA N T D ESIGN ................................................................................. 17 3.1 ISOLATING RING ARCHITECTURE .................................................................................................. 17 3.2 M OVEMENT DETECTION BY SOFTWARE ......................................................................................... 19 3.3 M OVEMENT DETECTION USING ACCELEROMETER ........................................................................ 19 4. POWER SAVING ELECTRONICS DESIGN................................................................... 21 4.1 DESCRIPTION OF THE BASIC CIRCUITRY......................................................................................... 21 4.2 POW ER BUDGET.................................................................................................................................22 5. THEORETICAL ANALYSIS OF FINGER PHOTOPLETHYSMOGRAPHY WITH R IN G SEN SO R ................................................................................................................................ 26 5.1 BACKGROUND....................................................................................................................................26 5.2 APPROACH ......................................................................................................................................... 27 5.3 M ODELING ......................................................................................................................................... 29 5.3.1 Optical Mo del............................................................................................................................. 29 5.3.2 Tissue Me chanical Mo del...........................................................................................................32 5.3.3 Dynam ics of the Arterial Wall ................................................................................................ 36 5.3.4 Parame ter Calibrationa nd Estimation. ................................................................................. 39 5.3.4.1 Validation of the Model and Determination of Stiffness of the Tissue by FEM.............39 5.3.4.2 Determ ination of Optical Properties : Capillaries and Veins .......................................... 41 5.3.5 Completing the Blood Vessel M odel....................................................................................... 46 5.4 VERIFICATION BY FINITE ELEMENT METHOD, NUMERICAL SIMULATION AND EXPERIMENT ..... 47 5.4.1 Analysis of the Sim ulation and Experiment.: Case 1.............................................................. 48 5.4.2 Analysis of the Simulation and Experiment: Case 2.............................................................. 51 5.5 VERIFICATION OF ADVANTAGES OF THE ISOLATING RING SENSOR BY FINITE ELEMENT METHOD ....53 6. SIGNAL PROCESSING WITH CORRELATION FUNCTIONS....... ......... 57 6.1 BACKGROUND....................................................................................................................................57 6 .2 TH E O R Y .............................................................................................................................................. 5 8 6.2.1 General De scription of the Signal Conditioning Process ..................................................... 58 6.2.2 Theoretical De scription of the Autocorrelation Function..................................................... 59 6.3 NUMERICAL SIM ULATION.................................................................................................................. 63 6.4 EXPERIMENT ...................................................................................................................................... 65 6.4.1 Experime nt Setup........................................................................................................................65 5 6.4.2 Experiment Results.....................................................................................................................67 7. PROTOTYPING AND FABRICATION ............................................................................ 68 7 .1 P A C K A G IN G ........................................................................................................................................ 6 8 7.2 ELECTRONIC COMPONENT SELECTION...........................................................................................69 7.3 THE POWER-OPTIMAL CLOCK FREQUENCY ................................................................................... 70 7.4 SOFTWARE DESIGN ............................................................................................................................ 72 7.4.1 Software for the M icroprocessoro n the Ring Side................................................................. 72 7.4.2 Software for the Host Computer with Artifact Detection ....................................................... 72 8. M INIATURIZATION ........................................................................................................... 74 8 .1 B A C K GR OU N D .................................................................................................................................... 74 8.2 ISSUES OF M INIATURIZATION ........................................................................................................ 75 8.2.1 How do we reduce size? ............................................................................................................. 75 8.2.2 What kind of circuit boards will we use? ............................................................................... 77 8.2.3 How do we reduce the power consumption from the viewpoint of hardware?........................ 77 8.3 PROCESS OF FABRICATION.................................................................................................................78 8.3.1 Finalize the circuit and collect the necessary components. ................................................... 78 8.3.2 Design a conductingp attern to be put on the ceramic substrate or the printed circuit board. .79 8.3.3 M ake the circuit board using gold as the conducting material............................................... 79 8.3.4 Put the components on the board and make connections........................................................ 79 8.3.5 Do external wirings and debugging ...................................................................................... 80 8.3.6 Software - In circuit Programming....................................................................................... 80 9. VERIFICATION AND BENCHM ARKING ...................................................................... 82 9.1 SOFTWARE-BASED ARTIFACT DETECTION ..................................................................................... 82 9.2 ADJUSTMENT OF INNER RING TENSION AND CONTACT PRESSURE............................................... 83 9.3 COMPARISON BETWEEN THE ISOLATING RING AND A NON-ISOLATING RING .............................. 85 9.4 BENCHMARKING ................................................................................................................................ 87 9.5 LONG-TERM M ONITORING EXPERIMENT........................................................................................... 91 9.6 DESIGN GUIDELINES FOR THE RING SENSOR ................................................................................... 92 10. CONCLUSIONS .................................................................................................................... 96 REFERENCES ................................................................................................................................ 98 6 List of Figures & Tables Figure 2-1 Conceptual diagram of the ring sensor Figure 3-1 Dislocation of ring sensors due to external load Figure 3-2 Construction of isolating ring Figure 3-3 Various Signals Detected by the Ring Figure 3-4 Detection of movement by 3-D accelerometer Figure 4-1 Block diagram of electronic circuit Figure 5-1 (a) Uncompressed finger under no external force (d=0). (b) Finger compressed by the ring due to an external force.(d>0) Figure 5-2 Optical model of the finger and optical elements. Blood vessels have different optical properties from the tissue. Figure 5-3 (a) Initial state of the ring with LED and photodetector (b) When the finger moves in the ring (Finger tissue is deformed.) Figure 5-4 Geometry of the LED, the photodetector, artery 1, and the skin capillary layer Figure 5-5 Change of arterial wall radius (Rr) with transmural pressure (Pt) Figure 5-6 Pressure-diameter relationship of human digital artery Figure 5-7 Pressure-diameter relation curve described by sigmoid function Figure 5-8 Finite element analysis of the finger segment under constant pressure of 120 mmHg. Figure 5-9 Deformed finger segment under external pressure, generated by FEM. Figure 5-10 Volumetric changes of the finger segment with various values for Young's modulus. Figure 5-11 FEM Analysis: Single Ring: E=20000, v=0.49, disp=2mm, angle=90 deg Figure 5-12 FEM Analysis: Single Ring : E=20000, v=0.49, disp=2mm, angle=20 deg Figure 5-13 Single Ring : E=20000, v=0.49, disp=2mm, angle=90 deg, 3-D View Figure 5-14 Single Ring : E=20000, v=0.49, disp=2mm, angle=90 deg, Cross Sectional View at x=0 Figure 5-15 Simultaneous recordings of chamber pressure (Pch), AC components of impedance variation (AZ), and DC components of the impedance variation (Z ). 0 Figure 5-16 Numerical simulation result of the volume change (AC components) of 1 cm finger segment under external pressure Figure 5-17 Numerical simulation result of the volume change (DC components) of 1 cm finger segment under external pressure Figure 5-18 Two cases of finger movements in the ring Figure 5-19 Visualization of movement at case 1 Figure 5-20 Single Ring - 3-D View : E=20000, v=0.49, disp=2mm, angle=20 deg Figure 5-21 Single Ring : E=20000, v=0.49, disp=2mm, angle=20 deg, Cross sectional view at x=0 Figure 5-22 Pressure distribution as a function of angle in the finger cross section generated by the lumped parameter model. Figure 5-23 Pressure distribution as a function of angle in the finger cross section generated by finite element method. Figure 5-24 Photoplethysmography and pressure at sensor unit from experiment in case 1 Figure 5-25 Photoplethysmography and pressure at sensor unit from numerical simulation in case 1 Figure 5-26 Pressure distribution as a function of angle in the finger cross section generated by the lumped parameter model. Figure 5-27 Pressure distribution as a function of angle in the finger cross section generated by finite element method. Figure 5-28 Visualization of movement at case 2. Figure 5-29 Photoplethysmography and pressure at sensor unit from experiment in case 2 Figure 5-30 Photoplethysmography and pressure at sensor unit from numerical simulation in case 2 7 Figure 5-31 FEM Analysis: Isolating Ring: E=20000, v=0.49, disp=2mm, angle=90 deg Figure 5-32 FEM Analysis : Isolating Ring : E=20000, v=0.49, disp=2mm, angle=20 deg Figure 5-33 Isolating Ring: E=20000, v=0.49, disp=2mm, angle=90 deg, 3-D View Figure 5-34 Isolating Ring: E=20000, v=0.49, disp=2mm, angle=90 deg, Cross Sectional View at x=0 Figure 5-35 Isolating Ring : E=20000, v=0.49, disp=2mm, angle=90 deg, Cross Sectional View at the Outer Ring Figure 5-36 Isolating Ring : E=20000, v=0.49, disp=2mm, angle=90 deg, Longitudinal View Figure 5-37 Isolating Ring : E=20000, v=0.49, disp=2mm, angle=20 deg, 3-D View Figure 5-38 Isolating Ring: E=20000, v=0.49, disp=2mm, angle=20 deg, Cross Sectional View at x=0 Figure 5-39 Isolating Ring : E=20000, v=0.49, disp=2mm, angle=20 deg, Cross Sectional View at the Outer Ring Figure 5-40 Isolating Ring: E=20000, v=0.49, disp=2mm, angle=20 deg, Longitudinal Cut-Plane View Figure 5-41 Isolating Ring : E=20000, v=0.49, disp=2mm, angle=90 deg, Wide Ring, 3-D View Figure 5-42 Isolating Ring : E=20000, v=0.49, disp=2mm, angle=90 deg, Wide Ring, Cross Sectional View at x=0 Figure 6-1 Signal Processing Flow Chart Figure 6-2 Periodic Source with 1 Hz Frequency s[n] Figure 6-3 Random Noise d[n] Figure 6-4 Combined Signal x[n] Figure 6-5 Autocorrelation function of x[n] Figure 6-6 Second autocorrelation function of x[n] Figure 6-7 Experiment Setup Figure 6-8 Original ring sensor signal and the result after autocorrelation Figure 6-9 Autocorrelation function showing the second peak of the heartbeat Figure 7-1 Isolating ring sensor designed for motion artifact minimization Figure 7-2 Comparison of Power Budget Figure 8-1 First Prototype Ring Sensor Figure 8-2 Ring Sensor with One or Two Circuit Boards Figure 8-3 Wire Bonding Machine Figure 8-4 Dimensions of Components Figure 8-5 Pictures of the miniaturized circuit boards Figure 8-6 Miniaturized ring sensors Figure 9-1 Signal contaminated by motion artifact Figure 9-2 Signal contaminated by ambient light influence. Figure 9-3 Experiment of pulsation amplitude and skin contact pressure Figure 9-4 Experiment of tension-strain characteristics of inner ring band Figure 9-5 Static force experiment Figure 9-6 Comparison between the ring sensor of the single body design and the isolating ring sensor under external static force. Figure 9-7 Comparison between the ring sensor of the single body design and the isolating ring sensor under acceleration. Figure 9-8 No external static force with contact pressure of 75 mmHg. Figure 9-9 Static force experiment with 75 mmHg contact pressure. Figure 9-10 Static force experiment with 11 mmHg contact pressure. Figure 9-11 Heart rate monitored by EKG, Fingertip PPG device, ring sensor Figure 9-12 A part of the two-hour monitoring test result Figure 9-13 Guidelines for the design of the ring sensor. Table 9-1 RMS error (beats/min) of the heart rates from the ring sensor compared with those from the EKG and fingertip PPG device 8 1. INTRODUCTION 1.1 Background and Objectives Ambulatory patient care makes up the bulk of medical care and affords the best opportunity for preventive medicine. The renaissance of interest in ambulatory care in general, and for the hi-risk cardiac patient in particular, is gaining ever-increasing momentum. With the aid of modem technology and a better understanding of physiological processes, medical care is experiencing a rapid evolution in both diagnostics and therapeutics. This technical progress now provides the potential for improved care of the patient in the ambulatory environment. From the administrative and economic aspects, moving the focus of care from the hospital to the ambulatory environment can bring about considerable economic benefits. From the patients' viewpoint, they can avoid being confined in the hospital environment solely to guard against any future outbreak, but still protected from those possibilities by online observation [1]. Vital sign monitoring is becoming increasingly important for securing independent lives as the population of aged people increases. Online, continuous monitoring allows us to detect emergencies and abrupt changes in the patient's condition. Especially for cardiac patients, online, long-term monitoring plays a pivotal role. It provides critical information for long-term assessment and preventive diagnosis for which long-term trends and signal patterns are of special importance. Such trends and patterns can hardly be identified by traditional examinations. Those cardiac problems that occur frequently during normal daily activities may disappear the moment the patient is hospitalized, causing diagnostic difficulties and consequently possible therapeutic errors. In this sense, continuous and ambulatory monitoring systems are needed to detect the traits. Although there have been many ambulatory monitoring systems developed and discussed, none of them have reached the level that can fully cover the patients' everyday lives, mostly due to the fact that those devices are bulky and inconvenient to carry. When taking a shower, for example, people tend to remove any ambulatory monitoring devices. Bathrooms, however, are one of the most dangerous places in the home. More than 10,000 people, mostly hypertensives and the elderly, die in bathrooms every year. Therefore, it is important that the ambulatory monitoring devices should be easy to wear in everyday life. Such long-term, ambulatory devices must be compact, lightweight, 9 and comfortable to wear at all times. They must be designed for low power consumption for long term use. Furthermore, they must be able to detect signals reliably and stably in the face of motion artifact and various disturbances. Unlike traditional monitoring systems, these devices are used under no supervision of clinicians. Data is collected from the daily lives of patients in an unstructured environment. The goal of this thesis is to develop the technology for obtaining reliable measurements of vital signs for long-term use. A miniaturized photoplethysmograph (PPG) device in a ring configuration will be designed and tested, which will be used for continuously measuring valuable information such as heart rate. It will be shown that the device meets diverse and conflicting requirements, including compactness, motion artifact reduction, minimum loading effects, and low battery power consumption. Mathematical modeling of the finger under external forces will help understand the underlying principle of vital sign monitoring by the ring sensor. Its benchmarking tests with the FDA approved PPG and EKG will show the validity of the technology. 1.2 Prior Work in the Field The ambulatory ECG (Holter) device, one of the most widely accepted ambulatory monitoring systems, was developed and extensively studied by N.J. Holter [2]. Bellet also devised a continuous 2-hour tape recording system using a similar device [3]. When the ambulatory ECG device was first introduced, the device was not immediately widespread due to concerns over the lack of previous documentation of coronary artery disease, the device's reliance upon T-wave changes, and the lack of recorder fidelity [1]. After many improvements and validity tests, the ambulatory ECG technology has gained increasing popularity. The ambulatory ECG, however, is not applicable to long-term monitoring for a period of several weeks or months. The machine is bulky, heavy, and uncomfortable to wear due to cumbersome wires and patches. Recently, a variety of vital sign sensors have been developed that are compact and easy to wear. Yamashita, et al. [4] attempted to develop a simple telemetry device for monitoring the pulse at a finger. Wristwatch-type pulse oximetry and blood pressure sensors have been developed and commercialized by several companies including Casio (BP-100 and JP200W-1V) and Omron (HEM-608 and HEM-609). These devices, although much easier to wear, have not yet been used clinically. Many technical issues still need to be solved prior to clinical use. In general, long-term, ambulatory monitoring 10