Expert Pacing System - FAU · fit. They are embedded in remote control systems com - municating with the implant in predefined time intervals and operate on programmed parameters

336 October 2000

Progress in Biomedical Research

Introduction

Cardiac pacemakers and ICDs are becoming increas-ingly sophisticated thanks to the progress in our elec-trophysiological and hemodynamic understanding ofthe heart and improvements in hardware and softwaretechnology. From the physician’s point of view, thegrowing range of pacemaker types and the expandingamount of statistical functions and stored data in theimplanted units increases complexity of the device andrequires more time and training to manually achieve anoptimal therapy. Hence, one of the main goals of cur-rent research and development activities is the simpli-fication of device adjustment: The implant should soonbecome a self-explaining system with a high degree ofautomation in the follow-up.In view of the development of future devices, the suc-cess of new and sophisticated therapeutic pacing con-cepts depends on their realization inside a self-learning

active system that automatically adapts its functions tothe patient’s specific conditions while detecting andforeseeing critical cardiovascular states and applyingthe most appropriate of all possible therapies. Expert-based systems will play a key role on the way of reach-ing these targets. They can be applied in all areaswhere important decisions must be made effectivelyand reliably. The expert in these systems is a softwaremodule that contains stored knowledge and tries tosolve specific problems in much the same way that ahuman expert would do. The expert systems can main-ly be useful in two different ways: • Decision support, e.g. by reminding the experienced

cardiologist of options or issues to consider, by mak-ing suggestions, by offering alternatives, by listingsimilar patient cases and solutions, and by givingdetailed documentation.

Expert Pacing System

M. SCHALDACHDepartment of Biomedical Engineering, University of Erlangen-Nuremberg, Erlangen, Germany

Summary

Within the last decade, knowledge-based computer systems have received a great deal of attention. Such expert sys-tems are being increasingly used in medical applications and have proved to be helpful tools in solving an impres-sive array of problems in a variety of fields. This paper gives an insight into the challenges and chances that expertsystems offer in the field of cardiac pacing systems. There are three types of integrated knowledge-based systems,depending on the kind of their interaction with the implant. Type 1 expert systems are used before or during implan-tation. They work on a large base of patients, medications and pacemaker/ICD data, possibly collected over yearsor decades. The knowledge acquired from these databases and from additional expert experience is used in orderto optimize the pacemaker indication process and to find out the optimal pacing settings considering the patient’sdisease and physical condition. Type 2 expert systems interact directly with or within the working implant. Theycan be realized as implant embedded systems aiming at solving a specific, well-defined problem. External PC-based expert systems are mainly used during follow-ups for analyzing interrogated Holter data of the implant. Bygiving suggestions on how to change pacemaker parameters and by discovering possible mismatches they may beof a great support for the practitioner. Type 3 expert systems offer the greatest potential benefit. They are embed-ded in remote control systems communicating with the implant in predefined time intervals and operate on pro-grammed parameters extracted from intracardiac signals.

Key Words

Pacing system, knowledge-based system, expert-based system, case-based system

October 2000 337


effects and side effects, targets, etc.) and pacemakers(types, modes, settings, targets, etc.). Case-based clas-sification systems collect as many patient case histo-ries as possible using additional knowledge to comparedifferent cases. Rule-based classifications represent themost common way of human reasoning and are formu-lated by simple if-then rules.The mathematical formalization of knowledge insidethe expert system is based upon these, shortly charac-terized, models of human reasoning. Thus, differenttypes of expert systems differ in how the knowledgebase is structured and how the inference engine works.One can distinguish between rule-based, frame-based,model-based and case-based systems or neural net-works (see [1] for details). The question now ariseshow to apply these types of expert knowledge systemsto the field of cardiac pacing systems.

Basic Strategies for Using Expert Knowledge inPacing Systems

The cardiologists’ expert knowledge is, beyond doubt,of prime importance in all phases of clinical patientcare concerning diagnosis and therapy strategies. Bothdecision data (e.g. symptoms, risk factors, diagnosticfindings, costs etc.) and possible decisions have to beidentified in these distinct phases. Therefore expertpacing systems have to support cardiologists’ reason-ing within multiple fields: diagnosis of basic disease,staging of basic disease, prediction of disease progres-

• Decision making, i.e. realized as an embedded spe-cialized module that use secure knowledge-based onverified research results in order to automaticallysolve specific well-defined tasks.

The first one is the most common in the field of cardi-ology till now. The process of decision finding can beillustrated using the overall structure of an expert sys-tem (Figure 1). The inference engine as the main tech-nical part of the system works on the specific taskusing a reasoning mechanism that draws conclusionsfrom the given data base and the knowledge base. Thelatter is a collection of experts’ knowledge of the par-ticular domain that the expert system is dedicated to.The quality of the collected knowledge is essential fora good performance of the system. That’s the reasonwhy the process of knowledge acquisition plays animportant role in developing expert systems andrequires an intensive cooperation between physiciansand knowledge engineers.The models of human reasoning show how differentkinds of knowledge should be stored in order to classi-fy patients. Logic decisions are usually represented inform of decision trees or tables (e.g. a special symptomleads to an accurate decision of how to solve the prob-lem). Heuristics are often formulated in form of exten-sively corroborated rules of thumb (the backgroundknowledge is of no interest). Statistical classificationsare mainly based on large representative data basesstoring the information on patients (clinical symptoms,ECGs, medication, diagnosis, etc.), drugs (dosage,

Figure 1. Structure of an expert system.

338 October 2000


sion and possible aftereffects, prevention of disease,risk stratification and deductions considering the fol-lowing drug and pacemaker therapy strategies. A gener-al view of an expert pacing system (Figure 2) let us dif-ferentiate between three types of integrated knowledge-based systems dependent on the kind of their interactionwith the implant.Type 1 expert systems are used before or during implan-tation. They work on a large base of patients, medica-tions and pacemaker/ICD data, possibly collected overyears or decades. The knowledge acquired from thesedatabases and from additional expert experience is usedin order to optimize the pacemaker indication processand to find out the optimal pacing settings consideringthe patient’s disease and physical condition.Type 2 expert systems interact directly with or withinthe working implant. They can be realized as implantembedded systems aiming at solving a specific, well-

defined problem. External PC-based expert systems aremainly used during follow-ups for analyzing interrogat-ed Holter data of the implant. By giving suggestions onhow to change pacemaker parameters and by discover-ing possible mismatches they may be of a great supportfor the practitioner.Type 3 expert systems offer the greatest potential bene-fit. They are embedded in remote control systems com-municating with the implant in predefined time intervalsand operate on programmed parameters extracted fromintracardiac signals.The complexity of an expert pacing system increasesfrom a Type 1 to a Type 3 expert system. The realizationof each expert system type involves many challengingtasks as well as promising opportunities to improveclinical practice. However, this requires close co-opera-tion between cardiology clinicians and knowledge engi-neers toward achieving expanding knowledge.

Figure 2. Expert pacing system in a general view.

October 2000 339


tions for pacing such as AVB, SSS, bradycardia, tachy-cardia, AF, etc., the distribution of pacemaker systemsand modes, the distribution of etiologies, etc [2]. Suchdata are usually presented in a statistical manner usingraw numbers and their percentage distribution or matri-ces combining different variables (Table 1). Other data-bases contain similarly structured information aboutdrug therapy or appropriate settings of pacemaker para-meters (AV delay, base rate, mode switch rate, thresh-olds, lead configuration, etc.).For example, further valuable databases are offeredby the industry standard Ann Arbor electrogramlibraries that provide benchmark wideband unipolarand bipolar intracardiac electrograms of cardiacarrhythmias for scientific investigation. These record-ings allow scientists and ICD developers alike to uti-lize identical data for device design [3]. Each record-ing consists of surface ECGs and intracardiac bipolarand unipolar electrograms of diverse cardiac arrhyth-mias (Figure 3).It is the knowledge engineer’s task to build knowledgebases out of all these statistical data. In cooperationwith a number of experienced physicians the engineercan create rule-based knowledge, e.g. by formulatingmore or less complex rules (combining any desiredinput parameters) like

IF (ECG-Indication = SSS & AVB) THEN (pacingmode = DDD(R)) is a good decision.

Statistic and Case-Based Expert Systems

Powerful expert systems can be used before the im-plantation of the ICD or the pacemaker, accessingvaluable information from large data pools or casebases (Type 1 systems). Pools containing a large num-ber of documented implantation data are of special interest, e.g. the databases of the German PacemakerRegistry offering detailed information on ECG indica-

Figure 3. Ann Arbor IEGM record of atrial fibrillation [3].

Table 1. ECG indications for, and pacing modes used in, ini-tial pacemaker implants in new federal states of Germany(all numbers in %; the numbers are taken from [2]).

340 October 2000


IF (ECG-Indication = AF) AND (COUNTER forAES = HIGH) THEN (pacing mode = overdrive)can be considered.IF (ECG-Indication = AF chronic) THEN (pacingmode = VVI) may be the best option.

or by simply considering the “probabilities” of statis-tics to guide possible decisions. Plausibility factorsrepresent the confidence one has that a fact is true or arule is valid. The additional knowledge of severalhuman experts in different fields can be combined togive a system more breadth and accuracy.Another way of establishing a knowledge-based sys-tem is based on the premise that human beings use

analogous or experiential reasoning. Therefore theengineer creates a case-based system, in which indi-vidual patient case histories are indexed according tothe presenting signs (ECG-strips) and symptoms, thelaboratory findings, the treatments (drugs and pace-maker settings) used, and/or any other factors that thephysician might later find interesting. The process ofcase-based reasoning is illustrated in Figure 4. Thesimilarity metrics in these systems are used to measurethe degree to which the given patient case matches pastcases. The adaptation module creates a solution eitherby modifying past solutions or creating a new solution.If no reasonably appropriate prior case is found then the

Figure 4. Case-based expert systems.

October 2000 341


tion of the patients’ status and of the functions of theimplant. A further aim is again to make suggestions forspecific changes of the programming parametersdeduced from Holter data, current settings and, ifapplicable, by considering the physical condition of thepatient. An additional benefit of these systems is thepermanent documentation of therapy including addi-tional graphical features on data and knowledge repre-sentation. For these purposes the expert system usesrule-based knowledge in order to analyze the availablediagnostic information, compare with limiting valuesor reference distributions, give hints on peculiar find-ings and make proposals regarding possible counter-measures. The practical value for the physician willdecisively depend on the completeness, correctness,practical relevance and other quality attributes of theimplemented clinical expertise, which is stored astables of rules. The physicians’ feedback on the results of the expertsystem plays an important role for further develop-ment. The main aim of the first clinical evaluations isto assess, complete, and extend the table of rules.Therefore such systems have to include a kind of learn-ing component giving the physician the possibility tocomplement and extend the base of knowledge. Thebroader the spectrum of the patients’ pacing indica-tions, programmed modes, age and other miscella-neous characteristics, the more meaningful the outputof the expert system is expected to be. Following thisway of verification of the rule tables and extending theknowledge base, the expert system grows up to a moreand more valuable tool during follow-up examinations.A first clinical evaluation study using such a systemled to very interesting and promising results [7]. The

physician has to find a new solution that can be added tothe case base, thus allowing the system to be improved.That way expert systems are established to support thephysician’s decisions on for example• the choice of pacemaker type;• the pacing mode to program;• the additional drug therapy;• whether standard pacing parameters should be

changed;• which parameter values may fit the patient’s condi-

tions best.The optimal settings for pacemakers offering advancedfeatures, e.g. prevention of arrhythmias by biatrial pac-ing, overdrive pacing, optimized AV-delay pacing areof special interest [4,5].An ideal Type 1 expert system asks the physician fordetailed data about the patient’s state and, after runningits knowledge and the physician’s input through itsinference engine, suggests a certain pacemaker pro-gram that can be downloaded from the programmer.Furthermore the system provides a detailed documen-tation on all data reflecting the particular patient caseand the process of decision making.

Local Expert Systems for Holter Data Analysis dur-ing Follow-up

Beyond doubt an automatic analysis by an external PCor laptop expert system (Type 2 system) can help thephysician while analyzing the diagnostic memory datainterrogated during follow-up examination of pace-maker patients (Figure 5). The diagnostic informationmemorized by the pacemaker and transferred to thecomputer is particularly helpful for a clinical evalua-

Figure 5. Analysis of Holter data during follow up by use of an external expert system [6].

342 October 2000


research group pointed out the potential clinical bene-fits through:• the full diagnostic potential provided by modern

pacemakers, which can be exploited without addi-tional effort;

• the detailed technical and medical expertise, offeredby knowledge-based support to the physician;

• the possibility of an early and precise detection ofthe progression of diseases and other risk factors;

• the expected long-term optimization of the therapy;• the consequent overall improvement of the quality

of follow-ups.Furthermore such automated systems can be regardedas the first step of remote control systems (Type 3expert systems).

Embedded Expert Systems for Specialized ProblemSolving

The computational power of microprocessors as wellas the memory capacity in currently used pacemakersand ICDs limit the range of algorithms possible to useinside the implant. For that reason, specialized and lesscomplex systems can be presently embedded to solvecertain well-defined tasks. Having in mind the greatamount of different tasks that the software of a pace-maker or an ICD has to do, a wide spectrum of possi-ble applications for expert systems is offered, above allregarding the fields of arrhythmia detection, arrhyth-mia prediction, arrhythmia prevention, arrhythmiaclassification, automatic optimization of implant para-

meters (AV delay, thresholds, etc.) or control parame-ters (stimulation rate, amplitude, duration, etc.). Theprocess of decision making by using available knowl-edge in these fields is reflected in form of easy decisiontrees and tables (clear logic) or by formulating simplerules and inference engines, building a fuzzy logic system. The implementation of the first idea is simpler andmore common because of less computational demands.For example, dual chamber ICDs operate with dis-crimination-algorithms in order to reliably discern theorigin of the tachycardia episodes. The SMART™(Biotronik, Germany) detection algorithm implement-ed in the Phylax AV (Biotronik, Germany) is based onthe concept that the chamber with the higher ratedenotes the origin of the tachycardia [8]. To differenti-ate in cases of equal heart rates in both chambers, addi-tional criteria are analyzed, namely sudden onset, ven-tricular and atrial rate stability, and regularity of the P-R intervals (Figure 6a). A successful discriminationleads to an appropriate therapy sequence (Figure 6b).Fuzzy systems are surely more complex to realize.However, they reflect human reasoning in a better waythan conventional methods do. Furthermore, theprogress of electrophysiological and hemodynamicunderstanding of the heart will require more intelligentand powerful algorithms in ICDs and pacemakers ofthe next generation.Figure 7 shows the structure of a fuzzy expert systemthat consists of a fuzzifier, inference engine and

Figure 6. SMARTTM algorithm for discrimination of tachycardia. Panel a: Decision tree. Panel b: Atrial and ventricular IEGMrecords in a successful discrimination of atrial flutter and effective termination by a 1-J shock [8].

a b

October 2000 343


ular arrhythmias [11,12]. It appears that the morpho-logical changes of this signal reflect the electrophysio-logic cellular processes and represent an indicator ofthe probability for the heart rhythm to become unsta-ble. Therefore the beat-to-beat parameters of this sig-nal play an important role for either risk stratificationof the patient or prediction of tachycardia. A morpho-logical alteration of this signal precedes observedepisodes of supraventricular arrhythmias in patientsfollowing aortic valve replacement [10]. Figure 8demonstrates the changes of MAP recorded from atri-al human epicardium prior to AF. In terms of predict-ing the onset of arrhythmias in this patient group, therate corrected MAP duration, the difference MAPd90-25 and the average change of cycle length andMAPd90 may all be used. The suggested predictionthreshold is depicted in Table 2. By creating a fuzzysystem using these parameters as input values, it ispossible to determine a risk factor for estimation of theprobability of tachycardia occurrence (Figure 9). Theadvantage of such a system is the integration of humanreasoning. Whereas conventional systems would acton exact values and fixed thresholds, fuzzy logic givesthe chance to interpret its input in a more human man-ner by defining fuzzy sets implemented as linguisticvariables. This procedure also considers the fact that itis often very difficult to set or find appropriate thresh-old values as they may differ between individualpatients. This is a typical fuzzy problem where the useof more or less heuristic rules is of great help. The dif-ficulty is surely the formulation of rules where furtherknowledge is to be acquired from future studies.Besides a change in the shape of monophasic actionpotential, other parameters are reported to be possiblepredictors or risk stratifiers, e.g. occurrence ofextrasystoles, heart rate variability, electrical alter-

defuzzifier. Fuzzification is a form of quantification inwhich the input data are assigned to so called member-ship functions. The latter give their input a specialdegree of membership to a fuzzy set using linguisticvariables that reflect human thinking. The inferenceengine translates the fuzzy input sets into fuzzy outputsets running the input through the if-then rule base ofthe system that was obtained by expert knowledge. Inthe end, the results of all rules are combined and thedefuzzifier derives an unfuzzy output. Interestingapplications of fuzzy systems concern again for exam-ple the classification of tachycardia for use in ICDs [9]or the estimation of stimulation rate in fuzzy-logic-based rate adaptive pacemakers with multiple sensorinformation [10].Regarding the field of arrhythmia prediction, clinicaland theoretical studies have shown that the monopha-sic action potential (MAP) may be used to predict theimminent onset of supraventricular as well as ventric-

Figure 7. Structure of a fuzzy system with multiple input andsingle output.

Figure 8. Atrial monophasic action potential prior to atrialfibrillation [10].

Table 2. Parameter limits for prediction of atrial fibrillationin patients following aortic valve replacement [10].

344 October 2000


the benefits of enhanced diagnostics [14]. Implants ofthe next generation comprise improved telemetry fea-tures and thus enable the establishment of home-basedoutpatient monitoring systems. They will be integratedparts of an enhanced service system, including variousgrades of services to patients and physicians. Besidesfinancial benefits this approach facilitates immensesupport for the physicians and thus optimum care fortheir patients. Patient monitoring will no longer berestricted to just the follow-up. Shorter monitoringcycles should result in early detection of situationsrequiring intervention by the attending physician,therefore contributing to enhanced patient safety andreduced subsequent costs. The main benefit of thesesystems is definitely the complete exhaustion of diag-nosis features that an implant can offer. Permanent andautomatic monitoring of the state of the patient bytransferring defined state variables will lead to furtherimprovements of diagnosis as well as enhanced thera-peutic strategies by discovering more efficient ways oftreatment. Considering the last aspect, large databasesthat are integrated in the service system offer a greathelp in analyzing the transferred data by the use ofexpert knowledge. Thus modern telecardiologyinvolves the possibility to use the advantages of pow-erful expert systems.

nants, heart rate turbulence, parameters using the ven-tricular evoked response (VER) and many others. Theidea is to choose a large set of those parameters, com-bine them by formulating simple rules and therebyestablish a diagnosis system for monitoring cardiacstate. This approach also includes a possible monitor-ing of medication [13]. Beyond doubt, this is a chal-lenging target that can only be reached step by step.The application of automated systems embedded in theimplant software that allow incorporation of human orexpert-knowledge-based decision making will surelyexpand. But the question arises whether technicalprogress allows an implant embedding of even morepowerful and more intelligent systems using deeperknowledge or whether it would be more suggestive toestablish them as external systems including the addi-tional benefits that large databases yield. The secondsolution seems to be very promising, above all consid-ering the immense advances in telecardiology.

Optimal Therapy Monitoring and Decision Supportby Expert-Based Remote Control

The rapid developments in the field of moderntelecommunications and wireless data transmissionwill surely enable physicians to take full advantage of

Figure 9. Fuzzy expert system for prediction of supraventricular tachycardia.

October 2000 345


during automatic measurements and enables the auto-matic transmission of rejection monitoring data to acentral station. The use of expert systems at this stagehelps to discover the significant correlation betweenthe results of endomyocardial biopsy and VER analy-sis. Furthermore, rule bases are established in order tooptimize the adjustment of medication dosage. Theprocess of knowledge acquisition is here of greatimportance, especially in the first phases of develop-ment. A central station for data storage and evaluationinvolves several advantages: patient histories enablecase-based comparison of different or similar situa-tions. Improved knowledge about how to change con-trol parameters can be embedded by adding newrules. Furthermore, the expert systems ensure a per-manent documentation of the ongoing therapy in thepatients. The deeper the knowledge and the more dataare stored in the service base the better the chances foran ongoing optimization of evaluation procedures.Following these tasks the “electronic biopsy” is aboutto become a viable alternative to the endomyocardialbiopsy, possessing advantages that include a lesserburden to the patient and overall lower costs to thehealth care system.Besides the expert-based use of telemonitoring oftransplant patients the same tasks have been set forenhancing the safety, quality of life, and service for allpatients with a pacemaker or defibrillator implant. Theidea of an ideal pacing system is to monitor the cardiacstate with an implanted pacemaker or ICD (Figure 11).

One important example, which is already in clinicaluse, relates to rejection monitoring after cardiac trans-plantation: the computerized heart acute rejectionmonitoring system (CHARM). Rejection monitoringprovides the possibility to weigh the dosage ofimmunosuppressive medication against the rejectionprocess. Within the CHARM project [15], non-inva-sive recording of intramyocardial electrograms is per-formed using fractal-coated epicardial leads and apacemaker with high-resolution intracardiac electro-cardiogram transmission (Physios CTM 01,Biotronik, Germany). The changes in the intracardiacelectrical signals are used to supplement the histolog-ically observed structural tissue changes in monitor-ing the rejection process (Figure 10a). The patient isexamined for rejection indications at the transplantcenter at predefined intervals. During measurements,the implant transmits the VER through a high-resolu-tion telemetric connection to an external data acquisi-tion system. Via the Internet, the data are sent to acentral data analysis workstation and evaluated auto-matically. The results are sent back to the physician atthe transplant center who decides whether and by howmuch the dose of medication should be changed orwhether to perform an endomyocardial biopsy(Figure 10b).At present, every rejection measurement requires thepatient to visit the transplant center. The next stage oftelecardiology development supports a home moni-toring system that allows the patient to remain home

Figure 10. Computerized heart acute rejection monitoring. Panel a: Typical averaged ventricular evoked response as obtainedfrom an intracardiac electrogram sequence during pacing. The rejection sensitive parameter (RSP) is calculated as the ampli-tude of a normal distribution, which was fitted to the terminal part of the repolarization phase using least square techniques[14,15]. Panel b: Overview of the system.

a b

346 October 2000


nal realization of the knowledge-based system is pre-ferred using the power of telecommunication. Thisincorporates the additional benefit of the access tolarge amount of data and the use of deeper knowledge.The implant receives the role of a diagnostic monitorobserving the cardiac function. It stores the measureddata and extracts the diagnostic markers. By applyingtelemetry equipment as indicated in Figure 12, thesemarkers are transferred to a center of knowledge via apatient base station. Here the state of the patient is per-manently stored, compared with historical patient dataand analyzed by the use of expert knowledge, includ-ing the possibility to monitor the success of pacing aswell as medication therapy. Evolutionary fuzzy algo-rithms (e.g. genetic algorithms) may enable self-learn-ing features of the system by, for instance, automaticevolving shapes and types of the membership func-tions as well as the fuzzy rule sets. The evaluated dataare summarized in a cardioreport that provides thephysician with detailed information on the state of thepatient and possible suggestions how to adjust the ther-apy.

For this purpose a set of defined electrophysiologicparameters would be extracted from the electrical sig-nals using enhanced features of implant electronicsand electrode sensors. The clinical variables takinginto consideration will depend on the functions of theimplant and encompass arrays of different eventrates, time delays reflecting conduction times, para-meters of variability, depolarization and repolariza-tion times, characteristic parameters of the MAP andVER signal as well as time constants that are evaluat-ed by system identification algorithms and reflectdynamic changes of cell behavior [11,12,13].Knowledge-based systems are established in cooper-ation with clinical experts. By use of fuzzy logicthese systems can combine their input variables in amanner of human reasoning and deliver control vari-ables for an ongoing optimization of pacemaker orICD therapy. Thus the implant can be seen as anadaptive control element that compensates forpathologies in an optimal way.As the expandable computational requirements of thementioned algorithms cannot be abandoned, an exter-

Figure 11. Observation and evaluation of cardiac state parameters using expert knowledge.

October 2000 347


of more profound electrophysiological and hemody-namic understanding of the heart, using intracardiacsignals, such as VER or MAP, as well as advances inthe fields of sensors, leads, batteries, microprocessors,memories, and signal processing techniques. Extraeffort will be required for extracting the valuable diag-nostic information out of the mass of measured data. The process of knowledge acquisition by interviewingphysicians, learning from past experiences, and unify-ing the know how of as many experts as possible willpromote the use of computational intelligence in car-diac applications. Knowledge accumulation leads tofurther progress in diagnosis and therapy of cardiacpatients resulting in new diagnostic information fromwhich the experts can acquire knowledge again. Thejustifiability of the expert system increases with deep-er knowledge leading to less rules and a less complexsystem. Keeping this point in view, there are chancesfor realization of intelligent knowledge-based systemsas embedded implant software modules in the nearfuture. Nevertheless, much work remains to be done toprovide patients and physicians with the full range ofextended facilities that expert systems promise.

Discussion

For medical tasks, the range of acceptable error issmall because of ethical concerns and malpracticerisks. Beyond doubt, however, expert systems in com-bination with the proceedings of modern telecardiolo-gy offer a wide spectrum of possible applications forpacing systems. Patient benefits include at firstarrhythmia prediction and prevention, disease detec-tion and classification, medication monitoring, rateadaptation, optimization of parameters leading toenhanced quality of life, risk stratification, and manyother more specialized functions.In view of computational efficiency it is preferable tosource out expert knowledge in the first phase ofdevelopment and establish external powerful knowledge-based systems as part of a multi servicecenter for the physicians. Additional benefits are pro-vided through the use of modern telecommunication.Also the application of automated systems embeddedin the implant software that allow incorporation ofhuman or expert-knowledge-based decision makingwill expand. Cardiac pacemakers and ICDs willbecome more and more sophisticated with the progress

Figure 12. Expert-based monitoring of cardiac state and therapy control by use of modern telecommunication.

348 October 2000


References

[1] Brown CE, O’Leary D. Introduction to artificial intelligenceand expert systems. 2000.http://www.bus.orst.edu/faculty/brownc/es_tutor/es_tutor.htm.

[2] Irnich W. German Pacemaker Registry; Statistics from thereport 1998. Associated to the Department of MedicalEngineering of the Justus-Liebig-University, Giesen,Germany.http://www.med.uni-giessen.de/technik/index-e.html.

[3] The Ann Arbor Electrogram Libraries. Ann Arbor MI, USA.http://www.electrogram.com/.

[4] Witte J, Reibis R, Baumann G, et al. Biatrial pacing for pre-vention of lone atrial fibrillation. Prog Biomed Res. 1998; 3:193-196.

[5] Attuel P. Suppression of atrial fibrillation using a new pacingalgorithm. Prog Biomed Res. 2000; 5: 13-18.

[6] Urbaszek A, Stoyan H, Schaldach M, et al. Expertensystemzur wissensbasierten Holterdatenanalyse. Biomed Technik.1998; 43 (Ergbd. 1): 346-347.

[7] Igidbashian D, Lonardi G, Barbiero M, et al. Clinical evalua-tion of pacemaker memory functions using a new computerbased expert system. Poster report. Cardiostim, 12thInternational Congress, June 14-17, 2000.

[8] Revishvili AS. Dual-chamber automatic implantable car-dioverter-defibrillator. Prog Biomed Res. 1998; 3: 9-13.

[9] Usher J, Duncan C, Jitu V, et al. A fuzzy logic controlledclassifier for use in implantable cardioverter defibrillators.PACE. 1999; 22: 183-186.

[10] Sugiura T, Kimura M, Mizushina S, et al. Cardiac pacing sys-tem based on fuzzy logic. Bulletin of the Research Institute ofElectronics, Shizuoka University. 1994; 29: 167-172.

[11] Pichlmaier AM, Lang V, Schaldach M, et al. Prediction of the onset of atrial fibrillation after cardiac surgery using the monophasic action potential. Prog Biomed Res. 1996; 1: 60-65.

[12] Weiss I. Analyse der elektrischen Funktionsmechanismenvon Herzrhythmusstörungen als Grundlage neuer Ansätze zurelektrotherapeutischen Prävention. Thesis, UniversityErlangen-Nuremberg, Germany, 1999.

[13] Merkely B, Lang V, Schaldach M, et al. Monitoring of anti-arrhythmic drugs with the monophasic action potential and theventricular evoked response. Prog Biomed Res. 1996; 1: 70-76.

[14] Schaldach M, Hutten H. Telecardiology – Optimizing thediagnostic and therapeutic efficacy of the next implant gener-ation. Prog Biomed Res. 1998; 3: 1-4.

[15] Bourge R, Schreier G, Hutten H, et al. A multicenter study onnoninvasive rejection monitoring of heart transplants. ProgBiomed Res. 1998; 3: 172-176.

ContactProf. Dr. Max SchaldachZentralinstitut für Biomedizinische TechnikFriedrich-Alexander-Universität Erlangen-NürnbergTurnstr. 5D-91054 ErlangenGermanyTelephone: +49 9131 85 22 805Fax: +49 9131 27 196E-mail: [email protected]

Documents

Expert Pacing System - FAU · fit. They are embedded in remote control systems com - municating with the implant in predefined time intervals and operate on programmed parameters