DigitalHuman (DH): An Integrative Mathematical Model Of .

DigitalHuman (DH): An Integrative Mathematical Model ofHuman PhysiologyRobert L. Hester, Ph.D.; Richard L. Summers, M.D.; Radu lIescu, M.D. Ph.D.; JoyeeEsters; and Thomas G. Coleman, Ph.D.Department of Physiology &Biophysics and Department of Emergency Medicine. University ofMississippi Medical Center. 2500 North State Street, Jackson, MS. 39216rhester@physiology.umsmed.eduAbstract. Mathematical models and simulation are important tools in discovering the key causalrelationships governing physiological processes and improving medical intervention when physiologicalcomplexity is a central issue. We have developed a model of integrative human physiology calledDigitalHuman (DH) consisting of -5000 variables modeling human physiology describing cardiovascular,renal, respiratory, endocrine, neural and metabolic physiology. Users can view time-dependent solutionsand interactively introduce perturbations by altering numerical parameters to investigate new hypotheses.The variables, parameters and quantitative relationships as well as all other model details are describedin XML text files. All aspects of the model, including the mathematical equations describing thephysiological processes are written in XML open source, text-readable files. Model structure is basedupon empirical data of physiological responses documented within the peer-reviewed literature. Themodel can be used to understand proposed physiological mechanisms and physiological interactions thatmay not be otherwise intUitively evident. Some of the current uses of this model include the analyses ofrenal control of blood pressure, the central role of the liver in creating and maintaining insulin resistance,and the mechanisms causing orthostatic hypotension in astronauts. Additionally the open source aspectof the modeling environment allows any investigator to add detailed descriptions of human physiology totest new concepts. The model accurately predicts both qualitative and more importantly quantitativechanges in clinically and experimentally observed responses. DigitalHuman provides scientists amodeling environment to understand the complex interactions of integrative physiology. This researchwas supported by.NIH HL 51971, NSF EPSCoR, and NASApressure regulation during orthostasis (8), butthese are also somewhat limited in theintegration of the entire body responses. In thispaper we describe a detailed integrative modelof human physiology, designed such that thephysiological descriptions can easily be changedby the user.1. 0 INTRODUCTIONMathematical simulations of physiologicalprocesses have become an important tool inunderstanding normal and pathophysiologicalprocesses within the body.Beard andcolleagues have presented a very detailedsimulation of cardiac metabolism, in particularenergy metabolism during cardiac ischemia (2).Secomb and colleagues have simulatedmicrocirculatory hemodynamics in vascularnetworks (10). Oxygen delivery to tissue hasbeen extensively modeled, (9) along withsimulations of VEGF release demonstrating thathypoxia-induced VEGF release is important todirect angiogenesis towards hypoxic tissue (7).There are an extensive number of publicationsin the literature describing mathematicalsimulations of individual organ systems, butnocomprehensivemodelstherearedemonstrating the integration across differentorgan systems. Olufsen et al. has describedintegrative models of the baroreflexlsympatheticnerve system interaction demonstrating bloodStarting with the Guyton cardiovascular model inthe late 1960's and continuing to the present,Guyton and Coleman demonstrated the use ofcomputer simulations for education purposesand to develop and test hypotheses concerningphysiological systems (1; 5; 6). Perhaps one ofthe best known historical models of integrativephysiology is the Coleman HUMAN model whichcontained a detailed description of circulatoryfunction (4).There are currently three available integrativemodels of the human body. All are based on theoriginal model "Human" written by Drs. JamesRandall and Thomas Coleman between 1981and 1987. The 1980 version of Human is129

available at Skidmore College and is used as aweb-basedteachingtool.QCP2005.exe(Quantitative Circulatory Physiology) developedat the University of Mississippi Medical Center(UMC) incorporates the cardiovascular, renal,respiratory, endocrine, and nervous systems. Amajor limitation of this model is that the programis written in C and is compiled. Parametervalues can be changed using slider bars andother active screen objects, but it is not possibleto change or add underlying equations.Therefore we have developed DigitalHuman(DH), a simulation of human physiology wherethe underlying physiological relationships areThe model is completelywritten in XML.specified in XML (which is both machine andhuman readable). This paper describes thebasics of the current version of DH.including the structure of the model, the controlof solutions and the display of results.The XML data files have several importantelements. With respect to the XML schema usedby DH, these are: variables which is used to declare (andsometimes define) variables. There are a varietyof variable types including ordinary, parameters,constants, timer variables and random variables. equations Declare and parameterizedifferential and implicit algebraic equations. functions Define curvilinear functions. definitions Create the blocks of math thatcalculate the derivative values and do additionalancillary math.One element that is important is the XMLelement curve . Sometimes, the preciseunderlying physiological description is notknown, but can be described as a curvilinearfunction. To accomplish this we have developedthe element curve which is described by thefollowing text demonstrating the relationshipbetween P0 2 and erythropoietin section.2.0 METHODSThe DigitalHuman simulation package iscomprised of a series of files that describe thephysiology (Structure files), a set of files thatdescribe the display characteristics of thesimulations (Display), and. the executableDigitalHuman.exe file. To solve and display theXML based physiological descriptions theexecutable DigitalHuman.exe file is a compiled(C ) code which consists of a fast XML parser,numerical methods used in solving algebraicand differential equations, and the code thatgenerates screen updates. No unusual librariesare used. The code is currently compiled forWindows using one of several Microsoft C compilers.2.1 XML Structure Files For DH the XML filesdescribing the physiological responses arelocated in a folder called "Structure". Within the"Structure" folder are a series of subfolders,each describing a particular organ or responses.Within these sub folders are simple text fileshaving a .DES extension that contain thephysiological descriptions. This organization isdesigned for ease of use in finding particularfiles. The individual XML files are simple textfiles and modifications can be made by anyoneusing a text editor.The physiological variables and relationships aredescribed using Extensible Markup Language(or XML). The extensible in XML's nameidentifies one of its major strengths as XML canbe customized to meet specific needs. In thecase of mathematical model documentation, wehave developed an XML schema that is used torepresent the details of mathematical models, curve name P02Effect /name point x 0.0 /x y 4.0 /y slope 0 /slope /point point x 35.0 /x y /y slope -0.14 /slope /point 0.0 point x 60.0 /x y -1.0 /y slope 0 /slope /point /curve The parser reads the text and fits the data to acubic spline. Since most data sets are small wehave determined that the model builder mustspecify the slope of the curve at each data pointto provide a better fit That is the approach usedin DH and the curve for the above equation isshown in Figure 1. The benefit of this method ofdescribing a curve is that the researcher doesnot have to do any mathematical analysis.130

IIVoriabIeIIlIockJ 9"'"l1O"'"-the document is logged and reported when theerror is detected.Parsing of DH requires 4 seconds on Delldesktop (2.8 GHz, 3 GB, XP). Subsequentparses require less time since Windows memorymaps recently read files. The quick re-parseisconvenientduringmodelresponsedevelopment when lots of parses are needed.DielTtWlIDT H2ODT-I(DT NaEPOSeae ·on.P02ElIedE . ecre60n .y3.0 RESULTS3.1 Physiological responses-1.O.X60.DH allows the user to adjust manycharacteristics of thepatient's physicalenvironment, from global conditions such asaltitude to local qualifiers such as temperature,humidity, and barometric pressure.Otherexternal parameters that can be adjustedinclude partial pressures of the individual gasesin inspired air and nutritional composition andamount of ingested food and fluids. Control ofthe patient's daily routine schedule allows theuser to adjust basic functions such as sleeping,working, and feeding on an hour-to-hour basis,whereas the Exercise Panel facilitates studyingthe effects of differing types of exercise on botha short and long-term scale.Total: 356ClooeFigure 1: Curve fitting capability of DHThe examples of the code and Figure 1describing erythropoietin secretion demonstratethe ease of writing code to describephysiological responses. The "curve" functionallows theinvestigator to describeaphysiological relationship with an arbitrarynumber of points (the minimum is two), alongwith the slope at each point. The numericalmethods then fit a cubic spline curve to the dataTheand use this fitted equation in DH.investigator can view the curve and determine ifthecurveaccuratelydemonstrates thephysiological responses.Various panels such as the Organ Details andBasic Physiology button groups allow the user toinvestigate and adjust physiological parameterson a more in-depth basis. Panel featuresinclude graphical data displays, informationbuttons, and adjustable variables. In someinstances, pathophysiological states can bemimicked through the use of radio and sliderbuttons that allow hormone levels to be clampedor fixed at a given level.2.2 Parser and SolverDH parses mathematical expressions directlyand evaluates them in its own math engine.There are some important rules for writing theXML code. A common source of error is thefailure to comply with these rules, however theparser will inform the user of any rules that arebroken so the code can be corrected. The XMLparser makes the following tests:In addition to manipulating physiologicalparameters, DH allows the user to administerpharmacological agents to treat a simulatedpatient. Currently, several drugs are availablefor interventional treatmentchlorothiazide,digoxin, furosemide, midodrine, insulin andepinephrine. Additional drugs can be added asthe quantitatively relationships regarding theireffect uponphysiologicalfunctionsaredetermined.Several additional treatmentoptions are available in the DH model, includingplacing the patient on a ventilator, administeringfluids via an IV drip, and performing a bloodtransfusion.1. Is the document well-formed? This meansdoes it conform to the general rules of XML Forexample, is each opening tag paired furtherdown with a corresponding closing tag?2. Is the document valid? Does it conform to itsXML defined document structure? Are all theelement names recognized?3. Is the data acceptable? Is text interpretable?Is the literal a valid number? Is the numberwithin range?The "Chart" panel provides graphical descriptionof blood pressure, body temperature, heart rate,and ventilatory rate. The full drop-down menuprovides additional windows with organ andDH uses a serial access (SAX) parser thatcompletes all three tests in one pass. The parserstops if an error is detected and the exact spot in131