Monitoring and Control of the Accumulation of Interstitial Fluid in the Legs by MELISSA BARBAGELATA S.B. Mechanical Engineering Massachusetts Institute of Technology June 2000 Submitted to the Department of Mechanical Engineering in Partial Fulfillment of the Requirements for the Degree of Masters of Science in Mechanical Engineering at the BARK Massachusetts Institute of Technology MASSACHUSETTS iT OF TECHNOLOGY September 2002 OCT 2 52002 C 2002 Massachusetts Institute of Technology LIBRARIES All rights reserved Signature of Author............................ r. --- I Melk sa Barbagelata Department of Mechanical Engineering . -,June 14, 2002 C ertified by............................... Hartihiko Harry Asada Ford Professor of Mechanical Engineering wji esis Supervisor Accepted by............. .............. Ain Sonin Professor of Mechanical Engineering Chairman, Department Committee on Graduate Students E Monitoring and Control of the Accumulation of Interstitial Fluid in the Legs by MELISSA BARBAGELATA Submitted to the Department of Mechanical Engineering on June 14, 2002 in Partial Fulfillment of the Requirements for the Degree of Masters of Science in Mechanical Engineering ABSTRACT Regardless of its physiological origin, leg edema is a troublesome condition for many people. Edema in the leg occurs due to an abnormal accumulation of interstitial fluid. This accumulation can be reduced through certain external physical interactions with the body. Currently, there is no method for optimizing the prescribed treatments necessary to reduce edema. Applying closed loop control to the physiological system allows for regulation of a relevant parameter, such as interstitial fluid. To accomplish closed loop control it is generally necessary to have a model of the system, to be able to control the input and to be able to measure significant output variables. A fully powered massage chair imposes mechanical actuation on the system by changing the hydrostatic pressure within the body. Experiments done using the chair limit the treatable subject group to those who experience edema are sitting down. While the mechanical actuation is applied, a bioimpedance sensor measures the total extracellular fluid across the legs, which includes the accumulation of interstitial fluid. However, impedance alone cannot differentiate between the changes due to shifts in venous pooling or due to shifts in the interstitial fluid. Therefore, a simplified physiological model specifically built to incorporate the effects of segmental fluid compartments is used to estimate the contribution to the impedance signal from each of these volumes. This work presents a new physiological model capable of estimating changes in interstitial fluid due to gravitational effects. Initial experiments conducted using the aforementioned massage chair and bioimpedance sensor have aided in both the tuning and validation of the model. The tuned model is capable of estimating the interstitial fluid contribution to the impedance signal, providing a means for the quantification and eventual control of edema. H. Harry Asada Thesis Supervisor: Title: Professor of Mechanical Engineering 2 Dedication Para Felipe 3 Acknowledgements As my six years at MIT come to an end, there are many people whom I would like to thank. I would first like to thank my graduate advisor Professor Harry Asada for allowing me to work under his tutelage. He has been supportive and patient throughout my research endeavors, helping me grow as a researcher. I would also like to thank him for giving me the opportunity to be his teaching assistant for 2.010 (Modeling, Dynamics, and Control III), which has been one of my most rewarding experiences while at MIT. The research presented here would not have been possible without close cooperation with Devin McCombie. His insight into the physiological system as well as his bi-compartmental model is much appreciated. Also special thanks to Lisa Allison for her help in some last minute experiments. Thanks to my lab mates in the d'Arbeloff lab who have helped me both by sharing their ample knowledge with me and by always being such kind friends. Reggie, Eric, Sam, Kyujin, Phil, Yi and Steve, I will treasure our lab lunches and lab outings. To all my friends who over the years have always been encouraging and supportive, especially during the hard times. Reshma, Anjli, Jeff, Nick, Varouj undergrad would have been a bore without you guys! Thank you, your friendship. Zach, Barbara and Lorraine, our friendship has endured a whole decade and I look forward to having you as friends for many more. All the new people I have met in grad school, it has been a pleasure to spend this last two years with you, in particular Danielle, Joaquin, Jenny and all the other HST people, who treated me like I was one of them. And of course Phil, who has not only been my biggest support but has become my dearest friend in the last two years. Thank you for being wonderful. Last but not least thank you to my family. Mami y Papi gracias por todo el apoyo que me han dado, no seria nadie sin ustedes. Fabrizio and Fressia I hope you guys follow your dreams, I love you very much. 4 Table of Contents Monitoring and Control of the Accumulation of Interstitial Fluid in the Legs .................. 1 2 M onitoring and Control of the A ccum ulation of............................................................. 2 Interstitial Fluid in the Legs ............................................................................................. 4 A cknow ledgem ents...................................................................................................... 5 Table of Contents.......................................................................................................... 7 Table of Figures ....................................................................................................... 7 Table of Tables ......................................................................................................... 8 Chapter 1 Introduction.................................................................................................. 8 1.1 Physiological M otivation................................................................................. 9 1.2 Closed loop Control........................................................................................ 13 Chapter 2 Physiology: Interstitial Fluid...................................................................... 13 Fluid Compartm ents...................................................................................... 2.1 15 Fluid shift w ithin a compartm ent........................................................................... 17 Fluid change across m embrane ............................................................................. 19 2.2 V enous Pooling............................................................................................. 19 2.3 Edem a ............................................................................................................... 19 Background ............................................................................................................... 20 Edem a dynam ics:................................................................................................... 23 Chapter 3 Sensors and actuations ............................................................................... 23 3.1 Sensors .............................................................................................................. 23 Bioimpedance Sensor............................................................................................. 28 H eart Rate Sensor ................................................................................................. 28 Tem perature Sensor ............................................................................................... 29 3.2 A ctuators ........................................................................................................... 29 Chair.......................................................................................................................... 30 Other ......................................................................................................................... 31 Chapter 4 Understanding the m odel.......................................................................... 32 4.1 H um an Body Observer ................................................................................. 32 partm ental m odel ............................................................................... 4.2 Bi-com 36 Sensitivity Analysis ........................................................................................ 4.3 41 Chapter 5 Implem entation .......................................................................................... 41 D ata Collection ............................................................................................. 5.1 41 5.2 Experim ental Setup........................................................................................ 43 Experim ental Protocol ................................................................................... 5.3 44 Chapter 6 Tuning the M odel...................................................................................... Chapter 7 A ctuation D ependent Response ................................................................. 50 50 Changing leg height ............................................................................................... 51 Com pression ............................................................................................................. Chapter 8 Conclusion ................................................................................................. 53 5 8.1 Conclusion about m odel ................................................................................... Future W ork ...................................................................................................... 8.2 References ......................................................................................................................... Appendix A ....................................................................................................................... Appendix B ....................................................................................................................... Appendix C ....................................................................................................................... 6 53 54 55 57 59 62 Table of Figures Figure 1-1: B asic Feedback Loop ................................................................................. Figure 2-1: Major Fluid Compartments and Membranes that separate these compartments. (modified from Guyton, 1996) ....................................................... Figure 2-2: Flow of fluid into and out of the capillaries............................................... Figure 2-3: Venous Pressures measured in the reclined and upright positions ............ Figure 2-4: Effect of hydrostatic and colloid osmotic pressure on fluid filtration/reabsorption balance ............................................................................................... Figure 3-1: Impedance for a cylinder............................................................................. Figure 3-2: Four electrode Impedance sensor. ............................................................. Figure 3-3: High and Low frequency impedance. (modified from De Lorenzo, 1997..... Figure 3-4: Resistance and Capacitance Transformation ............................................ Figure 3-5: MEW Real Pro Chair, shown with the legs extended................ Figure 4-1: Bi-compartmental segmentation ................................................................. Figure 4-2: Arterial System (McCombie, 2002)............................................................... Figure 4-3: Venous System (McCombie, 2002)............................................................... Figure 4-4: Time Response for output equations using bi-compartmental model......... Figure 5-1: Electrode arrangem ent .............................................................................. Figure 6-1: Time Response for fluid accumulation in the legs measured with B ioimpedenace signal............................................................................................ Figure 6-2: Time Response for fluid accumulation in the legs given by the bicom partm ental m odel............................................................................................. Figure 6-3: Comparison of impedance signal to model signal ...................................... Figure 6-4: Extracting components to bioimpedance signal......................................... Figure 7-1: Impedance response to leg position changes. ............................................. Figure 7-2: Effect of compression on fluid in the leg ................................................... 10 15 18 21 22 24 26 27 27 30 33 34 35 38 42 44 45 48 49 51 52 Table of Tables Table 1: Body Fluid Compartm ents .............................................................................. Table 2: State variables in bi-compartmental model ..................................................... Table 3: O utput Equations ............................................................................................ Table 4: Sensitivity results............................................................................................. Table 5: Arterial Blood Pressure measured on the Leg ................................................. Table 6: Lower Blood Pressure for the tuned and un-tuned parameters........................ Table 7: Physiological Model -- Independent Parameters............................................. Table 8: R esults from Sensitivity Study ........................................................................... Table 9: Tune independent parameters .......................................................................... 7 14 36 37 40 46 47 57 59 62 Chapter 1 1.1 Introduction Physiological Motivation The negative effects of certain physiological conditions could be minimized or prevented with proper intervention. The initial motivation for this project arose when considering the recent cases of death associated with deep vein thrombosis (DVT) in airplanes, otherwise referred to as economy class syndrome. [1] In these cases, a person has died because prolonged seating resulted in a thrombus, or blood clot, developing in the deep veins of the calf. [2] The clot formation may have been prevented had the person been able to walk around and kept the blood from stagnantly accumulating in their legs. While it is simple to prescribe that people take care to walk around the cabin during long flights, given the recent number of deaths [3] this method has not proven effective enough. A second condition whose symptoms can be reduced through external interaction with the body is edema. Here the motivation is based on an aging lady at a nursing home (or even at her own home). In certain situations, the lady may be placed in a chair for six hours at a time without much attention. At the end of this time, an accumulation of interstitial fluid may have developed in her legs. Although she may have various conditions that would make edema more severe, certain maneuvers or sitting positions could reduce the edema observed. 8 While both the lady at the nursing home and the man in the plane are capable of preventing or reducing their conditions, they do not. However, if instead of depending on their own actions they are able to depend on some external actuation to keep fluids, such as the interstitial fluid and the venous blood from accumulating in their legs, then there would be less possibility for problems to appear. This could be done through the use of closed loop control. 1.2 Closed loop Control Closed loop Control is a familiar concept for the human physiological system. In fact, the human body utilizes many different feedback loops to maintain internal homeostasis. Arguably, the most important feedback loops are those which regulate blood pressure, and those that deal with internal temperature control. While closed loop control within the body works seamlessly, attempting to integrate external closed loop control with the human body is not easy. Most interactions with the body employ open loop control. For example, a doctor prescribes a predetermined pillbased regimen for two weeks. No feedback information is gathered to determine whether the medicine is having the desired effect. Similarly, recent inventions such as the marionette bed [4] are able to turn patients without assistance, but are not able to determine when it is necessary to perform and action. The benefits of closed loop control are evident from experience in robotic design. Being able to feed back a desired value allows careful manipulation of the response. Figure 1-1 shows the key elements of a typical system with closed loop control. 9 Input Controller Process Feedback Output 1 .4 Figure 1-1: Basic Feedback Loop In this case, the process (often referred to as 'the plant') is the human physiological system, the controller will be the source of actuation, namely an automated chair, and the output to be fed back will be interstitial fluid indirectly measured with a bioimpedance sensor. Without feedback, control is based on estimated responses. For instance, reexamining the marionette bed, a nurse knows that it is important to turn a patient say every 4 hours. Though she may not know the exact state of the body during this period, she will still perform the prescribed action. The goal of this project is to set up a system that is able to monitor a physiological state in a person, focused on interstitial fluid in the legs. Since this desired state cannot be measured directly, a model must be set up to estimate the contribution of this variable to the measured signal. Then, actuation is used to maintain or impose a desired value for such physiological state. A system of closed loop control has been successfully accomplished in the rateresponsive pacemaker. [5] This device is able to monitor a variable such as a person's breathing rate and impose a heart rate that corresponds with such a breathing rate. The 10 pacemaker effectively controls the heart rate because it is implanted inside the body and acts directly on the heart muscle that determines the contraction of the ventricle. Many physiological states can be manipulated externally. For instance, diabetic patients can control the level of glucose in their blood by regulating how many sweets they eat, though much more sophisticated work has been done to automate this process. Chemical regulation of internal parameters is a very common way of imposing control upon the physiological system. Another way of imposing control is through thermal interaction. For example, the body's internal temperature can change depending on how many layers of clothing are used by a person. Finally, mechanical actuation can also affect certain states of the physiological system. Exercise can change the cardiac output, stroke volume and body temperature. Even postural changes can alter certain internal parameters. When designing closed loop control for the physiological system it is necessary to find a parameter that can be controlled through external actuation. Even more importantly, though perhaps not as limiting, as finding a parameter that can be non-invasively controlled, is finding one that can be non-invasively monitored. Non-invasive monitoring technology has made significant developments in recent history. Through technological advancements in ultrasound, photo-plethysmography, bioimpedance and others, many more physiological states can be measured non-invasively than before. Finally, in order to accomplish closed loop control once the appropriate sensors and actuators have been selected, the system to be controlled must be fully understood. To accomplish this a model is created that is able to estimate the characteristics that are relevant in the control of the chosen physiological states. Ideally, the model should run 11 parallel to the real system, and its predictions should match the observed physiological states. Before the model matches the real system, it may be necessary to tune the model parameters. 12 Chapter 2 Fluid Physiology: Interstitial It is important to understand the physiological characteristics that are relevant to the conditions that will be controlled. Since the focus will be on the control of interstitial fluid in the lower extremities and how it affects edema, the dynamics of fluid accumulation in the legs is most relevant. First, an anatomical description of the fluid distribution in the different compartments of the body is discussed. Then, the dynamics that give rise to edema are presented 2.1 Fluid Compartments Water makes up between 55 - 60% of the total body mass of the average adult. Even though this water exists everywhere in the body, it is can be separated into two major components. Water that lies inside the cells is referred to as intracellular fluid (ICF) and water that lies outside the cells is referred to as extracellular fluid (ECF). The extracellular fluid can be further divided into the water that exists in the plasma, the interstitial space, the fluid in the bones and other dense connective tissue, and transcellular fluid. Table 1 illustrates the percentage composition of each of these compartments. 13 Table 1: Body Fluid Compartments [61 Body Weight Total Body Volume [%] [Liters] Water [%] ICF 33 55 23 ECF 27 45 19 Plasma 4.5 7.5 3.2 Interstitial 12 20 8.4 Dense CT water 4.5 7.5 3.2 Bone water 4.5 7.5 3.2 Transcellular 1.5 2.5 1.5 60% 100% 42 liters TBW Even though dense CT water and bone water make up a fairly large percentage of the extracellular water, because the fluid moves very slowly within and across these spaces they can essentially be considered as constant, and unimportant when considering the effects of fluids flowing across membranes. On the other hand, the fluid in the plasma is constantly moving and interfacing with the fluid in the interstitial space. Therefore, these fluid compartments can be simplified into the major reservoirs, the intracellular fluid and the extracellular fluid consisting of the interstitial space and plasma. This distribution is shown in Figure 2-1. The interactions within and across these compartments will be further discussed. 14 ..... ------ ... . .. .. . .... .... .... Figure 2-1: Major Fluid Compartments and Membranes that separate these compartments. (modified from Guyton, 1996) [7] Fluid shift within a compartment The fluid compartments shown in the figure above are present over the entire body. When looking at a particular body segment there will be contributions from each of the compartments. For instance, when looking at a segment of the leg, there will be water found in the intracellular, interstitial and vessel compartments. Therefore, when studying the fluid shifts within a compartment, the shift actually occurs between different segments of the body, as will be discussed below. 15 PLASMA Plasma is the only fluid compartment that exists as a real collection of fluid in one location [6], namely in the blood vessels. This fluid is continuously pumped by the heart to all parts of the body. Normally an equal amount of blood is pumped into an area by arteries and carried back to the heart by the veins. However, certain postural changes will affect this balance. When a person stands up from a reclined position the effect of gravity causes a rapid fluid shift from the thoracic venous blood into the leg venous blood. This shift can be as large as 500 - 700 ml, which is then pooled in the veins of the leg [8]. Such drastic changes will result in an instantaneous reduction of cardiac output. However, an intact cardiovascular system should recover quickly. These changes in venous blood, though not as dramatic, are also evident when a person moves their legs from the downward position to a horizontal position and vice versa. Another factor that affects the volume accumulation in the legs is temperature. Temperature is particularly important when looking at fluid transfer in the peripheral system. In the legs, temperature drops will be related to the vasoconstriction of the blood vessels. Therefore, even if the core temperature of the body has not changed, the external temperature of the legs will have an effect on how much blood is being stored within the leg veins. The primary contribution to shifts in the plasma distribution is attributed to changes in fluid stored in the veins. This is because the capacitance of the veins is very large and can be changed to accommodate more fluid. The arteries, on the other hand, have a much 16 smaller capacitance. Therefore, the amount of fluid stored in the arteries will not change under normal conditions. INTERSTITIAL Since the interstitial fluid is not able to move as freely as plasma, fluid changes within the interstitial space are not as significant. Instead, fluid changes between the interstitium and plasma are of more importance. These are discussed in the following section. Fluid change across membrane During normal conditions the total fluid leaving the capillaries is equal to the fluid returned to the capillaries plus the fluid carried back to the circulatory system through the lymphatic system, as shown above. When this balance between influx and out flux is altered, there will be a change in the fluid accumulation. The two general causes of extracellular edema are: (1) abnormal leakage of fluid from the plasma to the interstitial space and (2) failure of the lymphatics to return fluid from interstitium back to the blood. [9] This discussion focuses on the leakage of fluid from the plasma to the interstitial space. 17 VENULE ARTERIOLE 0 0 CAPILLARY Less Fluid Returns Fluid Leaves Capillary Figure 2-2: Flow of fluid into and out of the capillaries The net transfer of fluid across the capillary wall is defined primarily by four forces. These forces combine in the following way to form the Starling hypothesis: F = K[(P- PJ)- (uz,- )z )] (Eq 1) Which states that the difference between capillary pressure, P, and interstitial fluid pressure, PI , minus the difference between plasma colloid osmotic pressure, 71 ,, and the interstitial fluid colloid osmotic pressure z,, scaled by the capillary filtration coefficient, K , which is the (the product of the specific permeability of the capillary times the surface area of the capillaries) determines the net flow into the interstitium, F . F is defined to be positive when there is a net outward flow of fluid from the capillary. Changes in any of these forces will affect the flow of fluid from the capillary. 18 2.2 Venous Pooling When a person changes from a reclined position to a standing position, blood stored in the veins of the chest and abdomen moves to the veins in the legs due to the change gravitational forces [8]. When this occurs the body responds as if it had just experienced a small hemorrhage including a drop in arterial pressure. If a person remains standing, the without movement the extra blood will remain in the legs. However, as a person begins to move this extra blood will reenter circulation, since the pumping action of the leg muscles will force fluid back towards the heart and out of the leg veins. Since the concern of these experiments is limited to changes that occur while sitting down, the pressure drops that may arise from standing and lead to conditions such as orthostatic hypotension [10] are ignored. Nevertheless, as the relative position of the legs to the heart changes while sitting down the volume of blood stored in the veins will also change. 2.3 Edema Background Edema refers to an abnormal build up of fluid in the tissues of the body. It can occur in several parts of the body including the lungs and the upper and lower extremities. When edema occurs in the legs it may have several causes. In the healthy individual, swelling occurs as a result of prolonged standing or sitting. [11] In addition, edema may be a sign of the presence of a serious disease. For instance, congestive heart failure will decrease the effectiveness of the heart's pumping, resulting in build up of fluid that is visible in the feet and ankles. Also, severe chronic lung disease will increase the pressure in the blood 19 vessels to the lungs. This pressure backs up to the right heart and ultimately to the veins emptying into the heart causing swelling in the feet and ankles. Since edema often indicates a serious condition, it is fundamental to understand the cause of edema when addressing its treatment. However, there are certain actions that can help the fluid in returning the veins and the lymphatic channels. [12] These actions will alleviate the discomfort and pain often associated with edema, even though they are not directly addressing the source of the problem. [13] Elevating the legs so that they are higher than the chest will help reduce the accumulation of fluid in the legs. Additionally, light exercise will cause a pumping action of the muscles. The compression of the legs will also help to keep the fluid inside the vessels. A common therapy to help alleviate edema while standing is the use of medial compressive stockings. [14] Edema dynamics: Edema occurs when there is an abnormal accumulation of fluid in the interstitial space. Looking back at Starling's Hypothesis, a positive value of F would cause an accumulation of fluid in the interstitial space. Many conditions can affect each of the factors contributing to flow across the capillaries. Bacterial infections, vitamin deficiency or burns can affect capillary permeability, Kf . Plasma Protein concentration, rp,, is decreased due to loss of proteins in urine and failure to produce proteins as is the case in liver disease. Capillary pressure, P, is influenced by excessive kidney retention of salt and water, high venous pressure, and decreased arteriolar resistance. [7] 20 HYDROSTATIC PRESSURE The factors mentioned above which disturb Starling's equilibrium all occur over longer periods of time. In addition, there is another transient factor that affects capillary pressure, namely gravity, see Figure 2-3. -42 2 +90 Venous Pressure [mmHg] 5 2 Figure 2-3: Venous Pressures measured in the reclined and upright positions[15] Gravity will have noticeable effects on the vascular pressures and volumes in various parts of the body, depending on the vertical distance of the column of blood from the heart. This increase in venous pressure will also raise the pressure in the venules and eventually in the arterial end of the capillary; this will result in an increase of filtration across the capillaries, and an increase in the accumulation of fluid in the interstitium [15], see Figure 2-4. 21 40 40 PA P OW 30 30 20 20 10 10 P b) Increased Venous Pressure a) Normal Figure 2-4: Effect of hydrostatic and colloid osmotic pressure on fluid filtration/re-absorption balance [151 Although there are many other factors aside from those listed that will affect the Starling balance, it is assumed that under normal conditions for P, A , and ,z will be operating at steady state. On the other hand, the transient effects of posture on P cannot be ignored; instead, they can be used to control the accumulation of fluid in the interstitial space. 22 Chapter 3 Sensors and actuations 3.1 Sensors Bioimpedance Sensor IMPEDANCE MODEL The principle of electrical-impedance is based on measuring electrical conductance to assess body volume changes. Bioimpedance has been used to measure many different types of biological electrical conductors, an ultimately to assess important physiological parameters. A few common ones include: total pulmonary function, cardiac output, segmental blood pulsation, and segmental venous reservoir. Bioimpedance is based on the fact that a small current can be safely introduced into the body. Then, by measuring the voltage imposed by the current between two points, it is possible to deduce the impedance of a particular area. For a cylinder, a simple model relating impedance to volume is shown below. 23 Z=p a z Z=p- (Eq 2) zA (b) (a) Figure 3-1: Impedance for a cylinder Where Z is the impedance measured across the cylinder, p is the resistivity of the matter, a is the cross sectional area, I is the length and V is the volume. If the length of a body segment remains constant the expansion of a blood vessel results in an increase in its cross sectional area and therefore volume [16]. This change in volume can be measured using bioimpedance. The volume change is modeled as an additional impedance in parallel to the first. Assuming that the measured length is constant, the change in volume is due to a change in the vessel's cross sectional area. The following was derived by Nyboer [16]. 24 AZ=Zi-Z2 (Eq3) =p_)012 V AZ - - 2 P1 V2 AV V1V2 plAV ZAV V2 V AZ- -AV Z V (Eq 4) Thus, the fractional change in impedance is inversely related to the change in volume. In other words, as the volume increases, the impedance decreases and its inverse, the conductance, increases. SENSOR MECHANICS The Quantum X from RJL Systems, Clinton Township, MI was used for Bioelectrical Impedance Analysis (BIA). The Quantum X is a four-electrode sensor. This means that the current flows through the two outer electrodes and the voltage is measured at the two inner ones, see Figure 3-2. Two-electrode configurations are also possible. However, when the voltage is measured at the same location the current is applied the quality of the resulting signal is poor. 25 ........... ... Smiusaidal constant acrrent source AC vector referece Resstance 0 degree s Reactance 90 dgre I D~ei~ Selectroe Current swrce electrode Current source electrode Measured biological resistance and reactance Figure 3-2: Four electrode Impedance sensor. [171 The Quantum X operates at a frequency of 50 kHz and inputs a current of 800 uA into the body. The voltage drop between the two inner electrodes is measured through an input impedance amplifier. At this frequency the measured impedance has contributions from both resistive and reactive elements. The resistive component is due to the extracellular fluid, while the reactive contribution is due intracellular fluid, since the cells in the body act like capacitors because of their cell membrane. At low frequencies, the current is unable to penetrate the cells. As the frequency increases though, some of the current is able to pass through the cells. At very high frequencies, all the current would go through the cells [18]. See Figure 3-3. 26 High Frequency Current (3)ID Cell Membrane Intra-Cellular Water Extra-Cellular Water Low Frequency Current Figure 3-3: High and Low frequency impedance. (modified from De Lorenzo, 1997 At a frequency of 50 kHz, measurements of the phase angle between the input and the resulting voltage can separate the resistive and reactive components of the response. The Quantum X, computes the reactance and resistance as if they existed as a simple resistor and capacitor in series. Since the corresponding biological model is of a resistor and capacitor in parallel, the data acquired must be converted into its parallel equivalent through the following transformation: Resistance and Capacitance in Series Rpa, = Rseries + cr Resistance and Capacitance in Parallel X~ Xc,par =Xcsre,+ =Xc,series R2 XR,,,,xc eie Xc,series Figure 3-4: Resistance and Capacitance Transformation SEGMENTAL MEASUREMENTS For full body impedance measurements using the Quantum X the electrodes are placed at the wrists and at the foot creating a full loop of current throughout the body. However, 27 since the focus of this project is the volume accumulation across the legs, the electrodes were placed at the foot and immediately below the knee. Such a method of segmental impedance measurements will be more accurate when estimating extracellular fluid in a confined area than the full body method. [19] In addition, because the geometry of this area is simple, it can be approximated with the simple cylindrical model described above. The resistive component to the impedance will include contributions from each of the fluids that make up the extracellular space as shown in Table 1. However, since only the interstitial fluid and vessel plasma will respond to dynamic changes the other contributions can be considered as adding to the DC component of the impedance signal but not to the changes seen. Heart Rate Sensor The heart rate sensor used is the Nellcor N-395. It is able to measure both the heart rate and oxygen saturation at the fingertip. It uses photo-plethysmography to methods to calculate the heart rate. Alternatively, the Ring Sensor [20], can also be used to noninvasively measure the heart rate. Temperature Sensor A simple thermistor is calibrated and used to record the temperature of the foot during experiments. These external temperature measurements can reflect the degree vasoconstriction or vasodilation in the leg. Since, as described earlier, both of these affect the amount of venous pooling. 28 3.2 Actuators Chair In order to control on a physiological state actuation must be able to affect the state of the body variables. Chapter 2 shows how both interstitial fluid accumulation and venous pooling are affected by a change in hydrostatic pressure. Therefore, by externally changing the hydrostatic pressure it is possible to change the states of both of these parameters. The Real Pro massage chair form Matsushita Electric Works, Ltd (MEW) was utilized for conducting experiments. The Real Pro chair is capable of conducting a series of automated, mechanical massage routines, which include the compression of the lower calves and ankle area. All of the components of the chair are fully powered, including the leg rest, which has a range of movement from-5 0 to 850 measured from a line perpendicular to the ground. It shall be noted that, the calf and ankle compression can be performed independent of all other actions. 29 Figure 3-5: MEW Real Pro Chair, shown with the legs extended Other Besides mechanical actuation, several other forms of actuation were considered. In particular, when considering fluid accumulation in the legs, passive electrical stimulation of the leg muscles would stimulate some of the fluid pooled in the veins to be sent back to the heart, through the pumping action of the muscles. Electrical stimulators can be placed over the leg muscles and have been approved by the FDA to improve blood flow. [21] 30 Chapter 4 Understanding the model The resistive signal that the bioimpedance sensor outputs includes contributions to the extracellular fluid from plasma in the blood vessels, interstitial fluid and other fluids. However, as previously discussed in Chapter 2, it is the interstitial fluid and the venous blood which will be most likely have large variations. As a result, the extracellular fluid can be modeled as being composed of only venous blood and interstitial fluid. Then, the changes observed with the impedance signal will be due only to changes in these two compartments. When analyzing the effect on a particular physiological condition such as edema, it is important to separate the relative contributions from interstitial fluid and venous pooling. This can be done by taking advantage of the fact that the volumetric changes for each have inherently different time constants. On the one hand, fluid shifts within the venous system happen almost instantly. On the other hand, fluid movement into and out of the interstitium takes place over significantly longer time periods. By manipulating the hydrostatic force applied to the system, it is possible to separate the effects from each of the different fluidic volumes. In order to clearly differentiate between each of these, a model is created that incorporates the necessary parameters to define these state variables. 31 4.1 Human Body Observer The Human Observer is a major undertaking by McCombie and Asada to create a continuous real time model of the human physiological system. It is based on a two-part simulation model made from a continuous model which refers to a discrete model for parameter validation. The initial discrete model is based on the agent based Mori model [22], which itself is based on the Coleman model. This model contains 25 subsystems and is able to calculates hundreds parameters. It includes information from the circulatory, respiratory renal and hormonal systems. One of the purposes of the Human Observer is to be able to impose certain conditions on the model to see how the real body would react by observing how the output variables of the model change. For example, one of the modules in the Human Observer is an exercise stimulus. Using this module, it is possible to see the effect of exercise on heart rate, blood pressure and other variables. At the same time, actual data is collected from the human and can be used to tune and validate the model for different subjects. The human observer separates the body into 11 segments based on their location in the body and calculates the circulation for all the segments simultaneously while taking into account any additional influences due to physical location. [23] 4.2 Bi-compartmental model The bi-compartmental model is a reduced version of the full Human Observer. Instead of separating the body into eleven segments, it separates it into only two major segments. The upper compartment includes the contribution from the head, the upper extremities and the torso, while the lower compartment includes the circulation only in the legs. This 32 architecture is chosen specifically to be able to model the effects that position changes have on the dynamics of the lower extremities. In particular, the circulatory changes that affect interstitial fluid are incorporated into the model since it is expected that these will yield important information to help prevent edema. UPPER BODY (TRUNK, ARMS, & HEAD) LOWER BODY (LEGS) Figure 4-1: Bi-compartmental segmentation McCombie created the bi-compartmental model using bond graphs to independently represent the flow within the arterial and the venous systems. The two branches of the bond graph are connected by including the parameters calculated in the model of the microcirculation for each segment, as shown in the figures below. 33 ARTERIAL SYSTEM UPPER BODY (TRUNK, ARMS, & HEAD) ARTERIAL CAPACITANCE C UPPER BODY ARTERIAL RESISTANCE UPPER BODY AORTIC CAPACITANCE I SF | 0 SE R ARTERIAL RESISTANCE R LOWER BODY ARTERIAL CAPACITANCE LOWER BODY C - ISF 10 C J_ - 1 SE 0 ' SF I MICROCIRCULATION UPPER BODY GRAVITATIONAL PRESSURE ON UPPER BODY LEFT VENTRICLE FLOW SOURCE GRAVITATIONAL PRESSURE LOWER BODY MICROCIRCULATION LOWER BODY LOWER BODY (LEGS) Figure 4-2: Arterial System (McCombie, 2002) 34 VENOUS SYSTEM UPPER BODY (TRUNK4 ARMS, & HEAD); MICROCIRCULATION VENOUS CAPACITANCE UPPER BODY VEIN WALL DAMPING 0UPPER BODY R I VENOUS RESISTANCE UPPER BODY GRAVITATIONAL PRESSURE ON UPPER BODY VENA CAVA CAPACITANCE RESPIRATION SE 0 VEIN WALL DAUPING - 1 - R VENOUS RESISTANCE LOWER BODY R ' - R - 1 PRESSUR C. HEART VALVE RESISTANCE SE 1- RIGHT HEART CAPACITANCE GRAVITATIONAL PRESSURE LOWER BODY VEIN WALL DAMPING R VENOUS CAPACITANCE LOWER BODY C 0 Is F MICROCIRCULATION LOWER BODY LOWER BODY (LEGS) Figure 4-3: Venous System (McCombie, 2002) The bi-compartmental model contains nine state variables. Seven of the variables are associated with each of the capacitances shown in the bond graphs above, while the other two are found in the microcirculation that links the arterial and venous systems to both the upper and lower segments. The state variables of the model are shown in Table 2. 35 Table 2: State variables in bi-compartmental model State Variable Parameter name used in MATLAB Aortic volume Q1 Artery volume, lower Q2 Artery volume, upper Q3 Vena cava volume Q4 Vein volume, lower Q5 Vein volume, upper Q6 Interstitial volume Q7 Interstitial volume Q8 Volume of the right atrium Q9 MATLAB* is used to solve the differential equations of the bi-compartmental model. In order to obtain the time profile for each of the state variables above, thirty-four independent parameters and nine initial conditions are necessary. These parameters include: compliances which are calculated from known values, resistances which are estimated from a known range, and other measurable characteristics such as height of heart, inspiration rate, and heart rate. [23] All of these parameters are ultimately required to obtain an accurate time profile of the output equations. When performing the model simulation a specific value must be used for each parameter. The chosen values represent a typical value for each parameter; however, they have not yet been tuned to a particular person. The values used are given in Appendix A and will be hereafter referred to as the base values. 4.3 Sensitivity Analysis Since it is very difficult to keep track of so many parameters at once, a sensitivity analysis was performed in order to determine how each parameter affects the chosen 36 outputs. First, the output equations were specified to incorporate the measurements that are of most interest, as shown in the Table 3. Table 3: Output Equations Output Variables Aortic pressure Arterial pressure (lower) Output Equation [state variables] = (Qi/Cal), Cal is the aortic compliance = (Q2/Ca2), Ca2 is arterial compliance (lower body) Venous volume (lower compartment) = Q5x106 Interstitial fluid (lower compartment) = Q7x106; The four output functions are the aortic pressure, arterial pressure in the legs, the fluid shift in the legs and the interstitial fluid shift. Notice that the sum of the third and fourth output variables is the extracellular fluid in the lower compartment. This sum can be directly measured using the bioimpedance signal. Since these output equations yield a time profile response, it is necessary to extract certain characteristics that can be used to compare whether the response is accurate or not. The time profile for each of the state variables will be a straight line, unless some disturbance is introduced into the system. Here, the disturbance introduced is a step change in the position of the legs, which changes the hydrostatic pressure in the legs. The time response for each output can be seen in Figure 4-4. 37 Figure 4-4: Time Response for output equations using bi-compartmental model Notice that the output variable may respond in one of three distinct ways. The lower arterial blood pressure (Q2) and lower body venous volume (Q5) each show a step change in the output response due to the step disturbance. Consequently, the initial and final systolic and diastolic values are necessary to define the response. The aortic blood pressure (Qi) shows a transient response to the disturbance, making only the final value of the systolic and diastolic pressures and the depression (or 'dip') relevant. Finally, the lower interstitial volume (Q7) has a ramp response to the disturbance, making the slope the relevant characteristic for this output variable. With this information, it is possible to extract one, three or four values that will completely describe each of the output equations. These values will serve for comparison 38 in the sensitivity analysis. Since now, it is unnecessary to compare the entire time response. The sensitivity metric is based on the following equation: outputextracted value Eq 5. apparameterbase value Where 8 aPparameterase valuerepresents poutput extracted value how much the base parameter has been changed and represents how much extracted output value has change as a result of the change in base parameter. The result of this calculation reflects the influence of changes in the independent parameter on the chosen output variable. The complete set of calculations, as well as all the sensitivity values for all parameters are found in Appendix B. The closer the sensitivity metric is to zero, the smaller the influence of the independent parameter is on the output. After performing the sensitivity analysis, we find that it is reasonable to keep track of a much smaller number of parameters for each output variable. These parameters are shown in the following table. 39 Table 4: Sensitivity results "*"UaP as Output Variable Independent parameter Aortic pressure Stroke volume (SV) 0.93 Heart rate (bpm) 0.72 Arterial resistance (Rart2, Rart3) 0.44 Stroke volume (SV) 0.45 Heart rate (bpm) 0.35 Heart height 0.61 Blood density (rho) 0.51 Compliance veins (C:v) 0.99 Heart height 1.02 Blood density (rho) 0.85 Heart height 1.17 Blood density (rho) 0.99 Bulk Flow Resistance (lower body) capillary to interstitium (Rbf2) -0.70 Molar concentration of protein in the blood (Cb) -0.32 Arterial pressure (lower) Venous volume (lower comp.) Interstitial fluid (lower comp.) autaPbase The last column of the table shows the influence of each independent parameter on the output variable, the larger the number the greater the relation between the parameter and the output. Negative numbers indicate an inverse relationship. The parameters chosen have sensitivity measurements of higher than 0.30. 40 Chapter 5 Implementation Once the relationship between internal independent parameters and the output equations is established, the information can be used to analyze our experimental results. The initial set of experiments was chosen so that the characteristics observed in the model can also be seen in the experiment. This chapter will describe the experimental setup as well as the initial investigational results. 5.1 Data Collection The Quantum X used in the experiments was specially modified by RJLSystems to provide an analog output for both the resistance and reactance components of the impedance signal. The impedance analyzer displays a digital signal in the range of 1 50Q for the lower leg, with resolution of ±0. IQ. The equivalent analog signal has a voltage of 0.150 V. The analog signal is amplified, using an operational amplifer, before being sampled by a National Instruments DAQ card (6024E). The data collection is controlled through a simple GUI written using LabView, and later analyzed offline using MATLAB® version 6.1. 5.2 Experimental Setup The Quantum X bioimpedance sensor described in Chapter 3 measures the impedance imposed by the body across the distance established by the two inner electrodes. In body composition analysis, the electrodes are placed on the feet and wrists, so that the 41 . ......... impedance can be measured across the entire body. However, since our experiments focus on measuring the volume accumulated in the leg, a couple of different electrode arrangements are possible. The impedance could be measured across the entire leg, or it could be measured only across the lower leg, as shown in Figure 5-1. The blocks represent the electrode placement; the letter i shows where the current was introduced and the letter v shows where the voltage was measured. Figure 5-1: Electrode arrangement Both of these arrangements provide similar information, since the experiments focus on measuring fluid accumulation. In fact, since the area between the knee and the waist remains stationary, there are no dynamic changes associated with this volume. When each was tested, the primary difference was that the configuration on the left was easier 42 to position and more comfortable to use. Therefore, the knee-ankle arrangement was chosen. Care was taken to place the electrodes on the same location each time, as shown in the figure above. 5.3 Experimental Protocol As previously stated, the focus of the experiments is on temporally-based fluid accumulation in the legs. There are several factors, discussed in Chapter 2, that affect this volume, but for the initial experiments the focus is on isolating the effects due to gravitational changes, i.e. changes in hydrostatic pressure. It is expected that after a change in the leg position there will be an initial change in the measured fluid volume due to the fluid shift within the vessels and a secondary contribution due to fluid shift between the capillaries and interstitial space. [24] This is believed to be true since the interstitial fluid shift occurs over a longer period of time, with a slow time constant. Ideally, the subject tested will not have any abnormal interstitial fluid accumulation before beginning data collection. Therefore, it is important for the subject to have been in an active situation immediately before sitting in MEW's Real Pro Chair, i.e. the subject should not have previously been sitting elsewhere for a long period of time. Once seated, the legs are raised and extended. Approximately two minutes later, the legs are moved into the gravity dependent position. The impedance signal is measured for the next twenty minutes. Special care is taken not to move the legs during that time, nor to contract the leg muscles, since both of these will affect the impedance signal. 43 Chapter 6 Tuning the Model The initial set of experiments was aimed at tuning the model described in Chapter 4. Data was collected on a subject following the guidelines in the experimental protocol outlined in the previous chapter. In the first step of the data analysis, only the resistive contribution to the bioimpedance signal was plotted, as shown in Figure 6-1. 160Flat region 158 156 Rapid movement of blood rl 14venous E'154 0 a 152 - Linearly increasing fluid amount i 150. 0-148 E 146 144 142 I 200 I 400 I 600 I 800 I 1000 I I I 1200 1400 1600 time [seconds] Figure 6-1: Time Response for fluid accumulation in the legs measured with Bioimpedenace signal Notice in the figure above that the measured impedance signal is initially flat. After about 160 seconds, the legs are moved towards the downward position. At this point, 44 there is a very rapid drop in the impedance due to the venous blood moving into the legs. Over the next 30 minutes, there is a fairly constant accumulation of fluid, denoted by the decreasing impedance. The data acquired from the experiment can be compared with the data generated by the bi-compartmental model discussed in Chapter 4. The relevant output equation for this comparison is the sum of the interstitial fluid shift and the venous blood shift. The model simulation for this volume accumulation is shown below. Fluid Volume Lower Legs j9$uu 3850 3800 - 3750 ' 3700 3650 36003550 350-n 0 I 200 I 400 I 600 I I 800 1000 I 1200 I I I 1400 1600 1800 2000 time (sec) Figure 6-2: Time Response for fluid accumulation in the legs given by the bi-compartmental model The response shown in the figure above is obtained using base values for all of the independent parameters. In order for the simulation to provide information about the 45 subject from whom the data was taken, the parameter values in the model must be properly tuned. Parameter tuning requires additional measurements to be taken. Measuring the output variables established in Chapter 4 will make it possible to tune the parameters that have been shown to have a high influence on each output variable. The arterial blood pressure can be measured in the upper body as well as the lower body. Though this measurement is not taken continuously throughout the experiment, it does add information about the points that have been extracted from the model. The following table shows the pressure readings taken at the leg. Table 5: Arterial Blood Pressure measured on the Leg Systolic Diastolic Legs up 125 ±2 70± 2 Legs down 155 ±2 92± 2 The pressure measurements shown above were taken by a nurse at the feet using a cuff and stethoscope, the accuracy provided refers to the accuracy using this method repeatability. Each measurement was done several times to ensure accuracy and the averages are shown above. In addition, there are other parameters that can be directly measured and inputted into the model. The heart rate is measured continuously with a Nellcor sensor. Since there were no significant heart rate changes, the measured average of 68 beats per minute was used in the simulation. Also, the position of the heart and the legs were measured and used in the model, all of these values are shown in Appendix C. The following table shows the 46 output values for the lower blood pressure both for the pre-tuned parameters and for the tuned parameters. Table 6: Lower Blood Pressure for the tuned and un-tuned parameters Blood Pressure Measured Simulated Base parameters Simulated Tuned parameters Legs Up 125 91 115 Diastolic 70 57 82 Legs Down 155 186 162 92 152 128 Systolic Systolic Diastolic The information acquired from the blood pressure measurements and the other measured parameters allowed us to tune some of the parameters in the model, including, leg height, heart height, heart rate, and arterial resistance. However, in order to produce more numerically relevant information using the model, more parameters should be tuned. Additional information can be extracted from the impedance measurements. For instance, the total decrease in impedance over the entire experiment can be measured. With this information, more parameters can be tuned, including venous compliances, right heart compliance and bulk flow resistance. The result is a much better match between the signal measured by the impedance sensor and the signal simulated by the model. The following figure compares the impedance signal converted to a volume with the volume signal from the model. For a complete list of the tuned parameters, see Appendix C. 47 3800 3750- simulated signal , 3700- 3650- measured signal E - 0 3600- 3550- 3500' 0 500 1000 1500 2000 2500 time [seconds] Figure 6-3: Comparison of impedance signal to model signal Now that the model signal is able to provide an acceptable match for the measured data, it is possible to extract the contributions from the interstitial fluid and from the vein to the fluid volume in the lower legs, as shown in Figure 6-4. When determining levels of interstitial fluid accumulation that are dangerous for edema, the interstitial fluid shift, not the venous blood shift should be examined. 48 Fluid Volume Lower Legs 3750 - 3700 3650 E 3600 - 05 3550 35000 200 400 600 800 1000 1200 1400 1600 time (sec) Lower Body Interstitial Volume 3600 I I I I I I I I I I I I 1800 2000 I 3580 -J E 3560 3540 E .0 3520 3500 3480' 0 I 200 400 600 800 1000 1200 1400 I I 1600 1800 2000 1600 1800 2000 time (sec) Lower Body Venous Volume 1501i E 100 F E .5 I I I 200 400 600 50 800 1000 1200 1400 time (sec) Figure 6-4: Extracting components to bioimpedance signal 49 Chapter 7 Actuation Dependent Response So far, the information obtained from the model has been compared to data obtained from the subject following a specific protocol. In order for closed loop control to be possible, the response of the human subject to actuation imposed by the chair must be understood. Changing leg height In order to record the applied actuation the chair was modified by attaching a low friction, rotary potentiometer to the leg lifting mechanism. This sensor allows for simultaneous collection of data related to both leg position and impedance changes. Lowering and raising the position of the legs relative to the heart changes the bioimpedance signal. These changes primarily affect the venous blood in the vessels, which is evident from the initial fast response of the signal to the input. 50 - -_:_ -""' L = - -.- .--- ......... -. ..... -- - I I 140 Impedance Chair Angle 135 E -13 125F- 120 7 B2M W4 .f Time [s] 8 em 9WO 920 94 Figure 7-1: Impedance response to leg position changes. Figure 7-1 shows the resistance signal as well as the output from the potentiometer relating chair position signal. The close correlation between resistance and position shows that the response of venous pooling occurs at the same time as the input is applied. Therefore, the change in venous pooling can be considered instantaneous. Compression The other primary actuation that can be imposed by the MEW chair is sinusoidal compression of the lower leg from the calf to the ankle. The response to this actuation is shown below. 51 186 Change in baseline fluid volume 15 175 165 - 111111' 11110111 -L I I 1900 190 160 17M 176D lam5 19w Time Isecon1sl 3101 2001 Figure 7-2: Effect of compression on fluid in the leg The intermittent compression pumps fluid out of the calves, but some fluid returns when the compression is released. The interesting thing to note is that the final value of the impedance is higher at the end of a series of compression. This implies that the net amount of fluid in the leg was actually changed through compression. By understanding the effect of similar actuations, it may be possible to change the state of the body. 52 Chapter 8 8.1 Conclusion Conclusion about model The model is vital to the analysis of information acquired from the sensor. The simple model presented here is capable of displaying internal parameters that may not be readily available directly from the sensor. This linear model provides a good first order approximation for estimating the amount of interstitial fluid in the body from the impedance signal. Tuning the parameters of the model helps ensure better agreement between the model and the real data, the more parameters that can be tuned the better the model information. However, the amount of parameters that can be tuned is limited by how much data can be measured in the experiment and by how such data can be incorporated into the model. By increasing the number of sensors used, the model will be more accurate. It is important to keep in mind that if the model does not include information about a particular sensor used, then the extra measurements will add no new information. For example, in some of these experiments temperature measurements were taken at the feet. However, since the model does not yet incorporate any information about temperature, this information could not be used to tune any parameters. 53 8.2 Future Work The system presented to estimate interstitial fluid in the legs is relatively simple. Many assumptions have been made in order to simplify both the overall model of the system and the sensors required to render valuable information. Under the situation specified, the model is able to match the experimental setup well. However, the model currently does not account for several other important factors that affect the signal. The eventual goal is to be able to run the physical experiments and feed the information directly into the complete Human Observer model. In order for this to be viable, more data must be fed back into the model. These data must be collected through added sensors to the body. Heart rate monitoring sensors and especially temperature sensors are essential in the next step of accuracy. Beyond these, there are other sensors that can also be incorporated. Electro-myogram can measure the muscle activity of the legs and determine whether there is muscl-related pumping of the venous blood by the muscles. Continuous blood pressure would be also be a valuable parameter to have, however it is still not practical to measure arterial blood pressure continuously and non-invasively. The more parameters that are measured, the more complex the model must be in order to integrate all the information relevant to the system. 54 References [1] Scurr JH et al. "Frequency and prevention of symptomless deep-vein thrombosis in long-haul flights: a randomised trial." Lancet May 12 2001. [2] Eklof B et al. "Venous thromboembolism in association with prolonged air travel". Dermatologic surgery. 1996 Jul; 22(7). [3] http://www.worldroom.com/pages/health/plane facts.phtml [4] Basmajian, A., Blanco, E.E., and Asada, H.H., "The Marionette Bed: Automated Rolling and Repositioning of Bedridden Patients," IEEE Robotics and Automation Conference (submitted 2002) [5] http://www.medtronic.com/brady/patient/rateresponsive.html [6] Brandis, K. Fluid Physiology (www.qldanaesthesia.com) [7] Guyton, A., Hall, J., "Textbook of Medical Physiology," 9th ed., Philadelphia: W.B. Saunders Company, 1996. [8] Smith, J Kampine John. Circulatory Physiology - the essentials. Williams & Wilkins. Baltimore, 1984. [9] Guyton, A., Hall, J., Textbook of Medical Physiology, 9th ed., Philadelphia: W.B. Saunders Company, 1996. [10] Stewart JM. "Pooling in chronic orthostatic intolerance: arterial vasoconstrictive but not venous compliance defects." Circulation 2002 May 14;105(19):2274-81. [11] InteliHealth.com: Edema [12] http://www.iowaclinic.com/adam/ency/article/003104trt.shtml [13] Lippmann HI. "Edema Control, Physics and Physiology" Proc. Rudolf Virchow med. Soc. New York 1969; Vol 27;, pp 170 - 179. [14] Jonker MJ et al. "The Oedema-Protective Effect of Lycra@ Support Stockings." Dermatology 2001; 203:294-298. [15] Smith JJ, Kampine JP. Circulatory physiology Williams & Wilkins. Baltimore, 1984. [16] Nyboer, J Electrical Impedance Plethysmography. Charles Thomas. Springfiel, 1970. 55 [17] Liedtke, R. "Principles of Biolectrical Impedance Analysis". rjlsystems.com. April 1997. [18] De Lorenzo A, et al. "Predicting body cell mass with bioimpedance by using theoretical methods: a technological review." J. Appl Physiol. 82(5) 1542-1558. [19] Zhu, F et al. "Dynamics of segmental extracellular volumes during changes in body position by bioimpedance analysis." J. Appl Physiol. 85(2): 497-504, 1998 [20] Asada, H., Shaltis, P., Rhee, S., "Validation and Benchmarking of a High-Speed Modulation Design For Oxygen Saturation Measurement Using Photo Plethysmographic Ring Sensors," Progress Report 3-2, HAHC, 2001. [21] www.fda.gov [22] Barbagelata M, Asada H, Mori, T, Kitamura T, . "An Agent-Based Physiological Digital Human and Its Application towards a Health Enhancing Chair" Progress Report 3-2, HAHC, 2001 [23] McCombie D, Asada H. "An Integrative Circulatory System Model Using Foreground-Background Multi-Time Scale Simulation Environment" Progress Report 33, HAHC, 2002 [24] SchUtze H. Hildebrant W, Stegemann J. "The interstitial fluid content in working muscle modifies the cardiovascular response to exercise." Eur J Appl Phyisiol (1991) 62: 332-336. 56 Appendix A The following table shows each of the independent parameters used in the bicompartmental model. The table includes the parameter name used in the code, as well as a description, the units and the initial values used. These initial values shall be subsequently referred to as the base values Table 7: Physiological Model Name Units -- Independent Parameters Base Values Description PHYSICAL rho Kg/mA3 density of blood plasma yh yl yu meters meters meters vertical height of the heart vertical height of the lower body vertical height of the upper body 1030.0 E 1.5 0.3 1.5 E E E Aorta Artery, Lower Body Artery, Upper Body Vena Cava Vein, Lower Body Vein, Upper Body Interstitial fluid upper body Interstitial fluid lower body Right heat 9.OOE-09 5.OOE-10 5.OOE-10 2.75E-08 1.25E-08 1.25E-08 3.50E-05 6.50E-05 1.OOE-07 C C C C C C G G COMPLIANCES Cal Ca2 Ca3 Cv1 Cv2 Cv3 Cif2 Cif3 Cra mA3/Pa mA3/Pa mA3/Pa mA3/Pa mA3/Pa mA3/Pa mA3/Pa mA3/Pa mA3/Pa RESISTANCES Rvl 1 Rvl2 Rvl3 Pa/(mA3/s) Pa/(mA3/s) Pa/(mA3/s) vessel damping, vena cava vessel damping, vein, lower vessel damping, vein, upper 1.OOE+08 1.OOE+08 1.OOE+08 G G G Ra102 Ral03 Pa/(mA3/s) Pa/(mA3/s) vessel resistance, arterial, lower vessel resistance, arterial, upper 3.OOE+07 3.OOE+07 E E 57 -- - .1, I 'ill I-,-MAP I- 1W., - Rv102 Pa/(mA3/s) - - - - --. vessel resistance, venous, lower 1.OOE+06 E Base Values Name Units Description Rv103 Pa/(mA3/s) vessel resistance, venous, upper 1.OOE+06 E Rart2 Rart3 Rbf2 Rlf2 Rbf3 Rlf3 Pa/(mA3/s) Pa/(mA3/s) Pa/(mA3/s) lower body arteriole resistance upper body arteriole resistance Bulk Flow Resistance (lower body) capillary, interstitium Bulk Flow Resistance (lower body) capillary, lymph capillary Bulk Flow Resistance (upper body) capillary, interstitium Bulk Flow Resistance (upper body) capillary, lymph capillary 1.44E+08 1.44E+08 2.00E+1 1 5.OOE+10 2.00E+1 1 5.00E+10 B B G G G G mA3 aortic volume artery volume, lower artery volume, upper vena cava volume vein volume, lower vein volume, upper interstitial volume, lower 6 Liters in cubic meters interstitial volume, upper 4 Liters in cubic meters Intial volume of the right atrium 9.60E-05 4.OOE-06 4.OOE-06 3.67E-06 3.33E-06 3.33E-06 3.50E-03 6.50E-03 0.OOE+00 E E E E E E E E E bpm ipm ml heart rate respiration rate stroke volume 75 20 70 E E E molar concentration of protien in the blood moles of protien in the lower body interstitial fluid 2.71 E+00 9.45E-04 B B moles of protien in the upper body interstitial fluid 1.76E-03 B Pa/(mA3/s) Pa/(mA3/s) Pa/(mA3/s) INITIAL VOLUMES q10 mA3 q20 q30 q40 q50 q60 q70 q80 q90 mA3 mA3 mA3 mA3 mA3 mA3 mA3 OTHER bpm ipm sv BiComp -- Internal Cb mol/mA3 0.27 mol/mA3 0.27 mif2 mif3 mol/mA3 TCR Pa/(mA3/s) Total Capillary Resistance, 40% of avg. arteriole resistance 4.11E+07 B R502 Pa/(mA3/s) right heart valve resistance 1.OOE+07 G B E C G ----- Physology Book: Human Physiology, Fluid dynamics Estimated from known range Calculated from estimated paramters (E, mu, L, r) Guessed to fit the model 58 Appendix B The table below shows the results from the sensitivity analysis. The first column contains all the independent parameters used in the bi-compartmental model, while the first row contains extracted measurements from each of the relevant output parameters. The larger the number, the larger effect a change in the independent parameter will have on the output variable. The parameters that have the highest influence on each variable are shown in bold and red numbers to help in their identification. Table 8: Results from Sensitivity Study Parameter name SV Q2 jump 0.01 amount Qend Q2 beg Q2 end sys sys changed sys 0.46 0.93 0.94 0.50 0.93 0.74 0.01 0.00 0.08 0.06 0.56 0.40 0.00 0.00 0.23 0.15 0.64 0.35 0.07 0.06 0.39 0.00 0.22 0.06 0.33 0.01 0.19 0.01 0.00 0.00 0.00 0.00 0.00 -0.01 -0.01 0.00 0.00 0.00 0.00 0.00 0.08 0.00 0.00 0.00 0.00 0.02 0.00 0.01 0.00 0.00 0.00 -0.10 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.17 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.20 0.00 0.00 0.00 0.00 0.00 0.00 -0.80 -0.32 -0.61 -0.21 0.00 -0.09 0.00 -0.07 0.00 -0.06 HR -0.20 HR -0.60 0.72 0.61 0.35 0.09 Rart2,Rart3 TCR Ra102,Ra1O 1.00 1.00 1.00 0.44 0.12 0.10 0.50 0.14 -0.02 0.25 0.07 -0.01 Rv102,Rv 0 1.00 0.00 0.00 0.10 1.00 0.50 -0.40 0.20 0.20 0.50 0.50 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1 3 Rbif2 Cb Rbif2 leg ht. q80 q70 q80 q70 q80 q70 Q7 slope 0.21 0.44 0.36 0.10 0.60 3 05 jump 0.00 0.90 0.73 SV HR -762 Q5 end Q5 beg sys sys 0.59 0.09 59 .. - - 7- q20 q1O Parameter name - - - ...... ...... - -- - - 0.00 0.00 0.00 0.20 0.00 0.13 0.00 0.50 end Q2 beg Q2 amount Olend sys sys changed sys 0.00 -0.12 Q2 jump - - 1 . ................. .............. - 0.00 0.00 0.06 0.00 05 end Q5 beg sys sys 0.00 -0.01 05 jump 0.00 0.00 07 slope q40 q60 q4 q30 q4 q30 q30 qlO q60 q1 q q20 q20 q60 q50 q50 C: if RbI mIf2,mIf3 Rblf3 rho 0.20 0.20 0.50 0.50 1.00 0.20 1.00 0.20 1.00 1.00 1.00 1.00 0.50 0.50 0.50 0.20 1.00 0.50 1.00 0.10 2.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.14 0.00 0.15 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.51 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.14 0.00 -0.14 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.85 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.11 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 -0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.01 0.12 0.04 1.00 heart,upper ht -0.20 0.00 0.00 0.61 1.20 1.03 0.00 1.20 1.18 heart height -0.20 0.00 0.00 0.61 1.20 1.03 0.00 1.20 1.17 0.00 0.99 0.91 1.01 0.00 0.00 0.00 0.00 0.01 -0.01 0.00 0.01 0.00 0.00 0.00 0.00 -0.01 0.00 1.00 -0.01 -0.02 -0.01 -0.03 -0.09 0.01 0.92 0.00 0.00 0.00 0.00 0.00 0.00 1.01 0.11 0.00 0.00 0.00 -0.02 0.00 0.00 C: v 1.00 0.00 -0.01 0.00 ml2,mif3 Ipm Cra ipm Cra C: a C: a & v 1.00 0.50 9.00 -0.50 1.00 1.00 1.00 0.00 0.00 0.00 -0.01 -0.01 -0.09 -0.09 0.00 0.00 0.00 -0.01 -0.01 -0.09 -0.10 0.00 0.00 0.00 0.00 -0.01 -0.04 -0.05 The second column of the table above shows the amount by which the independent variable has been changed. This value is necessary to normalize the sensitivity for different parameter changes. The following example will show how the information was acquired for the shaded box. 60 First the independent parameter disturbed is the heart rate which was changed from a base value of 75 bpm to 60 bpm. This resulted in a change of -0.20 as shown in the second colunm. The systolic value for the aortic pressure was extracted both before the change in heart rate and after. calculated P -P aft"e' befo" Then the effect of the change on the pressure was . Finally the pressure change effect is divided by the effect of the before parameter change AEffectpressue and a normalized value is acquired. AEffectheart rate 61 .......... Appendix C The following table shows the values for the tuned independent parameters as compared to the original base parameters. These parameters simulate the responses shown in Chapter 6. The values in red bold numbers shown which parameters have been changed through the tuning procedure. Table 9: Tune independent parameters Base Tuned Name IDescription PHYSICAL rho density of blood plasma yh yl yu vertical height of the heart vertical height of the lower body vertical height of the upper body 1030.0 1057.0 1.5 0.3 1.5 0.85 0.25 0.85 9.OOE-09 9.OOE-09 5.OOE-10 5.00E-10 2.75E-08 1.25E-08 1.25E-08 3.50E-05 6.50E-05 1.00E-07 5.OOE-10 5.OOE-10 2.75E-08 6.25E-09 6.25E-09 3.50E-05 6.50E-05 0.00E+00 COMPLIANCES Cal Aorta Ca2 Ca3 Cvi Cv2 Cv3 Cif2 Cif3 Cra Artery, Lower Body Artery, Upper Body Vena Cava Vein, Lower Body Vein, Upper Body Interstitial fluid upper body Interstitial fluid lower body Right heat RESISTANCES Rvl1 Rvl2 Rvl3 vessel damping, vena cava vessel damping, vein, lower vessel damping, vein, upper 1.OOE+08 1.OOE+08 1.00E+08 1.OOE+08 1.OOE+08 1.OOE+08 RalO2 Ral03 Rv102 vessel resistance, arterial, lower vessel resistance, arterial, upper vessel resistance, venous, lower 3.OOE+07 3.OOE+07 1.00E+06 3.00E+07 3.OOE+07 1.OOE+06 62 Rv103 1.00E+06 vessel resistance, venous, upper Base Tuned Name Description Rart2 Rart3 Rbf2 Rlf2 Rbf3 Rlf3 1.00E+06 lower body arteriole resistance upper body arteriole resistance Bulk Flow Resistance (lower body) capillary, interstitium Bulk Flow Resistance (lower body) capillary, lymph capillary Bulk Flow Resistance (upper body) capillary, interstitium Bulk Flow Resistance (upper body) capillary, lymph capillary 1.44E+08 1.44E+08 2.00E+1 I 5.00E+10 2.00E+1 I 5.00E+10 O.OOE+00 O.OOE+00 4.00E+1 I 5.OOE+10 2.00E+1 1 5.OOE+10 9.60E-05 4.00E-06 4.OOE-06 3.67E-06 3.33E-06 3.33E-06 3.50E-03 6.50E-03 O.OOE+0O 0.00E+00 7.20E-05 4.OOE-06 6.66E-08 7.33E-06 1.67E-06 3.50E-03 6.50E-03 0.OOE+00 75 20 70 68 20 70 INITIAL VOLUMES q10 q20 q30 q40 q50 q60 q70 q80 q90 aortic volume artery volume, lower artery volume, upper vena cava volume vein volume, lower vein volume, upper interstitial volume, lower 6 Liters in cubic meters interstitial volume, upper 4 Liters in cubic meters Intial volume of the right atrium OTHER bpm ipm sv heart rate respiration rate stroke volume BiComp - Internal Cb mif2 mif3 molar concentration of protien in the blood moles of protien in the lower body interstitial fluid moles of protien in the upper body interstitial fluid 2.71 E+00 9.45E-04 1.76E-03 2.71 E+00 9.45E-04 1.76E-03 TCR Total Capillary Resistance, 40% of avg. arteriole resistance 4.11 E+07 4.11 E+07 R502 right heart valve resistance 1.00E+07 1.001E+07 63