Requests for medical advice from patients and families to health care providers who publish on the World Wide Web.

BACKGROUND: The Internet is a novel, rapidly growing means of worldwide public communication. METHODS: We reviewed all unsolicited electronic mail and other communications from nonmedical individuals requesting medical information over a 12-month period from the physician at 1 established site on the World Wide Web. This site was the only Internet site with a primary focus on cardiac arrhythmias. RESULTS: Seventy unsolicited inquiries were received from 39 patients and 20 family members (the sources of 11 inquiries are unknown) from 20 states, Washington, DC, and 9 foreign countries (locations of 15 inquiries are unknown). Follow-up was obtained in 22 cases. The inquiries concerned cardiological conditions in 67 cases (96%) and cardiac electrophysiologic conditions and procedures in 52 cases (74%). The goals of the inquiries were diagnosis (15), therapy (48), prognosis (1), and patient education (6). On follow-up of 22 cases, the people initiating the inquiries stated that they were reassured (16), consulted a general cardiologist (1), consulted a cardiac electrophysiologist (4), or visited a tertiary care electrophysiology center (1). CONCLUSIONS: The increasing use of the Internet by the general public seeking specific medical information for themselves and for their families suggests a widespread, unmet need for objective medical advice. This study demonstrates that the public can choose accurately whom to ask for subspecialty advice in the area of cardiovascular diseases. Professional societies and regulatory agencies should develop physician guidelines for providing medical advice over the Internet.

Validation of the EINTHOVEN model-based computerized electrocardiogram rhythm analysis system with three classes of clinical arrhythmias.

Rhythm analysis by commercial systems does not meet clinical needs well, because (1) differential diagnosis of complex rhythms is not performed, (2) common rhythms are often misdiagnosed, and (3) transitions between rhythms are not described. We have developed a model-based diagnostic software system named EINTHOVEN that is designed to address the above limitations. A demonstration is available on the World Wide Web at http:@einthoven.uokhsc.edu. The system has been validated using simple rhythms from introductory electrocardiogram (ECG) textbooks. We present here the results of evaluation with more complex rhythm strips taken from clinical records and intermediate-level ECG textbooks. Rhythm strips were described by the onset and offset of each electrical event (P wave, QRS complex, and T wave) and by a morphology classification for each event. The rhythms included a variety of supraventricular and ventricular rhythms. The analysis was considered correct if it named all correct diagnoses in a rhythm strip, incorrect if it completed the analysis and failed to name the correct diagnoses, and indeterminate if it failed to complete the analysis. The system was designed not to complete an analysis if it could not explain an entire rhythm by at least 1 pathophysiological model. The test rhythms were not used to develop the system. Forty-six of 56 test rhythms were diagnosed correctly, and 8 were not analyzed completely. The 2 incorrect diagnoses were atrial tachycardia with variable conduction (diagnosed as intermittent complete heart block) and atrial fibrillation (diagnosed as irregular junctional tachycardia). All 56 rhythms were diagnosed correctly after minor technical improvements to the system. The processing time of the system was 7.6-fold (range 1.5-to 16.9-fold) faster than the elapsed time of the individual records. These preliminary results suggest (1) that computer-based interpretation of complex rhythms is possible, (2) that further software development is necessary to reach a clinical level of accuracy, and (3) that there are no theoretical obstacles to achieving this goal.

An algorithm for complete enumeration of the mechanisms of supraventricular tachycardias that use multiple atrioventricular, AV nodal, and/or Mahaim pathways.

The EINTHOVEN system is a model-based expert system that interprets the cardiac rhythm from the electrocardiogram. It simulates the expected behavior of realistic semi-quantitative cardiac models constructed by heuristic rules to generate interpretations that include both text descriptions and event-by-event causal explanations in the form of ladder diagrams. The simulation has been limited by an inability to predict all possible behaviors of hearts with more than one reentrant circuit. We now describe an algorithm that overcomes this limitation. Its output has been validated by an independent possibility-tree analysis. Timing and storage measurements are presented for models with up to three slow atrioventricular nodal pathways, four atrioventricular pathways, and a single atriofascicular (Mahaim) pathway. This is the first report in the literature of an algorithm that enumerates all possible mechanisms for reentrant supraventricular tachycardias that use atrioventricular, atrioventricular nodal, and/or atriofascicular pathways in humans.

Expert system reasoning about dynamic systems by semi-quantitative simulation.

Current symbolic knowledge representation techniques are inadequate for describing complex dynamic systems, which are time-varying and contain feedback loops. These systems are particularly difficult to represent when quantitative knowledge about the relationships within the system is incomplete. This paper describes a symbolic extension of the system dynamics method which can answer 'what if' questions about system dynamics models semi-quantitatively. The method consists of (1) definition of generic model 'building blocks' which map directly onto difference equations; (2) development of a symbolic causal model of a system in terms of the generic functional 'building blocks', any available quantitative or semi-quantitative parameters, and a set of generic default values; (3) automatic translation of the model into a system of first-order difference equations; and (4) numerical integration of the equations by standard methods. A complex model with semi-quantitative parameters, representing the human cardiovascular system, is used to illustrate the method. This method couples the descriptive abilities of mathematics with the symbolic power of causal inference methods by providing the same knowledge base for both computational levels. It also eases truth maintenance, knowledge acquisition, and explanation. It may prove useful for expert system development in a variety of application domains whose processes are time-varying and homeostatic.

The EINTHOVEN system: toward an improved cardiac arrhythmia monitor.

Contemporary cardiac arrhythmia monitors, used commonly in intensive care settings, are highly sensitive to artifact, resulting in high false alarm rates, inability to detect P waves reliably, and crude rhythm interpretation. We report on two new approaches that address these problems: a noise preprocessor that characterizes the type and degree of artifact in an ECG, and a model-based rhythm interpretation algorithm.

Model-based interpretation of the ECG: a methodology for temporal and spatial reasoning.

A new software architecture for automatic interpretation of the electrocardiographic rhythm is presented. Using the hypothesize-and-test paradigm, a semiquantitative physiological model and production rule-based knowledge are combined to reason about time- and space-varying characteristics of complex heart rhythms. A prototype system implementing the methodology accepts a semiquantitative description of the onset and morphology of the P waves and QRS complexes that are observed in the body-surface electrocardiogram. A beat-by-beat explanation of the origin and consequences of each wave is produced. The output is in the standard cardiology laddergram format. The current prototype generates the full differential diagnosis of narrow-complex tachycardia and correctly diagnoses complex rhythms, such as atrioventricular (AV) nodal reentrant tachycardia with either hidden or visible P waves and varying degrees of AV block.

Model-based interpretation of the ECG: a methodology for temporal and spatial reasoning.

A new software architecture for automatic interpretation of the electrocardiogram is presented. Using the hypothesize-and-test paradigm, a semi-quantitative physiological model and production rule-based knowledge are combined to reason about time- and space-varying characteristics of complex heart rhythms. A prototype system implementing the methodology accepts a semi-quantitative description of the onset and morphology of the P waves and QRS complexes that are observed in the body-surface electrocardiogram. A beat-by-beat explanation of the origin and consequences of each wave is produced. The output is in the standard cardiology ladder diagram format. The current prototype can perform the full differential diagnosis of 2:1 atrioventricular (AV) block, and can handle correctly complex rhythms such as AV nodal reentrant tachycardia with either hidden or visible P waves, and varying degrees of AV block.

A-to-D conversion from paper records with a desktop scanner and a microcomputer.

A novel method for digitizing signals contained in paper records is presented. This method is based on the use of an inexpensive optical scanner to translate the image on paper into a binary, bit map data structure. Several algorithms which recognize the signal line in the bit map and translate it into a series of numbers which are equivalent to the output of electronic analog-to-digital converters are described. The method was validated by comparison both with idealized test patterns of varying frequency content and with electronically digitized pressure and pressure time derivative tracings from chronically instrumented dogs. The root mean square error for the physiological signals was 3.5-3.9% of peak-to-peak full scale, corresponding to roughly 50% more than the thickness of the signal line on the paper.

Digitization of electrocardiograms by desktop optical scanner.

The fidelity of a semiautomated technique for converting paper electrocardiogram (ECG) tracings to digital form by optical scanning was examined. Sample tracings from one nonmechanical and three mechanical ECG writers (recorders) were used. The optically scanned signals were compared with the digitized version (402 Hz, 12-bit precision) of the original analog signals using time- and frequency-domain correlation coefficients and root mean square error. A total of 261 QRS complexes and 207 RR intervals were examined in 21 leads acquired from 8 patients. When data were low-pass filtered at 25 Hz, the correlation coefficients for the 261 QRS complexes were 0.997 +/- 0.005 (mean +/- SD) for the time domain data, 0.992 +/- 0.010 for the complex frequency domain (amplitude and phase) data, and 0.998 +/- 0.002 for the power spectrum. The corresponding correlations for the 207 RR intervals were 0.993 +/- 0.008, 0.992 +/- 0.008, and 0.993 +/- 0.009. The RMS errors, normalized for signal amplitude, were 2.62 +/- 1.28 (percent +/- SD) for QRS complexes and 1.82 +/- 0.87 for RR intervals. The correlations for the mechanical ECG recorder tracings were the same or better than those of the nonmechanical recorder, and the RMS errors were generally smaller. When data were low-pass filtered at 105 Hz, the correlation coefficients ranged from 0.984 to 0.996 for the QRS complexes and 0.982 to 0.988 for RR intervals. Root mean square errors were 4.54 +/- 2.03 and 2.38 +/- 1.14, respectively. For purposes of arrhythmia analysis by QRS classification, digitization of ECG signals by optical scanning appears equivalent to acquisition via standard analog-to-digital conversion.

Automated analysis of intracardiac electrograms obtained during extrastimulus tests using a three-dimensional electrophysiology model.

Catheter ablation procedures are performed by highly trained and experienced cardiology subspecialists. Yet the massive amount of data produced during these procedures creates a data overload problem that can impede the performance of even the best practitioners. This may be evidenced by (1) overlooking important signal features, (2) misinterpreting the signals, and (3) misinterpreting catheter locations in the heart, all of which can lead to increased procedure duration, applications of radiofrequency energy to the wrong part of the heart, or both. This article presents the first results from a project aimed at developing a model-based system for interpreting intracardiac electrograms in near real time. The system is intended to assist physicians in interpreting the enormous amounts of data recorded during catheter ablation studies. It is an extension of the Einthoven system that has been extended to account for the three-dimensional relationships in the cardiac conduction system as recorded in the various intracardiac electrograms. The new three-dimensional cardiac conduction model and the enhancements to Einthoven's reasoning algorithms are presented. The locus of this study is on interpreting the results of ventricular extrastimulus tests. Data collected for this study and the output generated by the system are presented.

EINTHOVEN and tolerance for human error: design issues in a decision support system for cardiac arrhythmia interpretation.

Decision support systems are becoming increasingly accepted in medical practice in the United States. Clinicians recognize the need for aid in interpretation of complex cardiac rhythms. The EINTHOVEN system is being developed to meet that need. In this paper, we address the need to deal with errors in the input due to inaccuracies in hand annotations by the inexperienced user and to interact with the user to correct them. Four specific types of input errors are described: missing waves, mispositioned waves, mislabeled waves, and extra waves. General and specific mechanisms by which these errors can be recognized and remedied are described. These results may be interesting as an example of the practical problems that arise in the design of real-world expert systems.

Multihormonal induction of alpha 2u-globulin in an established rat hepatoma cell line.

A subclone of the FU5-5 rat hepatoma cell line has been isolated which is inducible more than several hundred fold for the 20,000 dalton form of the major rat urinary protein alpha 2u-globulin. The basal relative synthetic rate (RSR) in growth medium containing 10% fetal calf serum was less than 2 X 10(-6) of total protein synthesis. Both dexamethasone and insulin were necessary for induction, and yielded a maximum induced RSR of 4-8 X 10(-3). Triiodothyronine (T3), dihydrotestosterone (DHT), rat growth hormone (GH), and estrogen, all of which have been shown to influence the induction of alpha 2u-globulin in the intact rat, were without effect on the cell line. A factor present in fetal calf serum was also necessary for maximum induction, since dexamethasone plus insulin in serum-free medium raised the RSR to only 3 X 10(-5); exogenous T3, GH, and DHT could not substitute for this serum factor. The kinetics of induction by dexamethasone were slow, with a lag of approximately 48 hr followed by a period of increasing RSR for 6-20 days. Removal of dexamethasone from induced cells led to an exponential decline in the RSR (t 1/2 15 hr). The concentrations of dexamethasone and insulin that could yield half maximum induction were 5 X 10(-8)M and 3 X 10(-11)M, respectively. Higher concentrations of insulin, although still in physiological range (10(-9)M), inhibited induction. At yet higher insulin levels, beyond the physiological range, alpha 2u-globulin synthesis returned to maximum values. The lack of DHT, T3, and GH requirement for alpha 2u-globulin induction in this cell line may mean that a regulatory aberrancy has occurred in this transformed cell line, or, alternatively, that these hormones act indirectly in the intact animal. This cell line should prove useful for the study of the molecular events associated with alpha 2u-globulin induction and for genetic approaches to the problem of multihormonal regulation of gene expression.

Nonlinear heart model predicts range of heart rates for 2:1 swinging in pericardial effusion.

We analyze two mathematical models of Rigney and Goldberger (14) of heart swinging in large pericardial effusions. Both models represent the torques due to the outflow of blood from the heart. The first assumes that the duration of systole does not vary with heart rate (in beats/min), whereas the second assumes that it varies linearly with heart rate. We examine the motion of the heart for heart rates between 50 and 200 and for a range of initial positions and velocities. Both models predict that the heart swings once every other beat (2:1 swinging, giving rise to electrical alternans) in a discrete range of heart rates and swings once per beat otherwise; both models explain the appearance and disappearance of 2:1 swinging mathematically. The first model predicts a rate range from 105 to 116 for the occurrence of 2:1 swinging. The second model predicts the same qualitative behavior but with 2:1 swinging occurring at heart rates between 88 and 119, which agrees well with published clinical data showing 2:1 swinging at heart rates between 90 and 144. We describe an analysis program for ordinary differential equations that analyzed the models quickly and automatically.

Model-based rhythm analysis of the ECG. Evaluation of a prototype implementation.

Current computer algorithms that interpret cardiac rhythms based solely on the surface electrocardiogram are limited, yet offer many benefits to health care. To address the limitations, novel computer algorithms for the automatic diagnosis of complex cardiac rhythms based solely on the surface electrocardiogram are presented. Using the hypothesize-and-test paradigm, a physiologic model of the cardiac conduction system and production rule-based knowledge are combined to reason about the time- and space-varying characteristics of complex heart rhythms. In addition, an evaluation of a prototype implementation of the algorithms is presented. A database of the time of onset, width, and shape classifications of each P wave, QRS complex, and T wave from 59 electrocardiographic strips was developed from an introductory textbook by hand-annotation using calipers. The database was not used in the development of the prototype. The prototype's diagnoses were reviewed by a clinical cardiac electrophysiologist who was not involved in the development process. Pair-wise comparisons among the prototype, textbook, and cardiac electrophysiologist, assuming either the textbook or electrophysiologist as the gold standard, were performed. The specific comparisons performed were prototype versus textbook, electrophysiologist versus textbook, prototype versus electrophysiologist, and textbook versus electrophysiologist. For all diagnostic categories, sensitivities of 88.0%, 97.2%, 78.6%, and 82.1%, respectively, and specificities of 99.2%, 98.5%, 99.7%, and 99.8%, respectively, were attained. When accounting for design and implementation limitations of the prototype, sensitivities of 93.0%, 98.5%, 89.1%, and 92.7%, respectively, and specificities of 99.4%, 99.2%, 99.6%, and 99.8%, respectively, were attained. The results indicate that these algorithms offer clinical advantages over currently available arrhythmia analysis systems.

A client/server system for remote diagnosis of cardiac arrhythmias.

Health care practitioners are often faced with the task of interpreting complex heart rhythms from electrocardiograms (ECGs) produced by 12-lead ECG machines, ambulatory (Holter) monitoring systems, and intensive-care unit monitors. Usually, the practitioner caring for the patient does not have specialized training in cardiology or in ECG interpretation; and commercial programs that interpret 12-lead ECGs have been well-documented in the medical literature to perform poorly at analyzing cardiac rhythm. We believe that a system capable of providing comprehensive ECG interpretation as well as access to online consultations will be beneficial to the health care system. We hypothesized that we could develop a client-server based telemedicine system capable of providing access to (1) an on-line knowledge-based system for remote diagnosis of cardiac arrhythmias and (2) an on-line cardiologist for real-time interactive consultation using readily available resources on the Internet. Furthermore, we hypothesized that Macintosh and Microsoft Windows-based personal computers running an X server could function as the delivery platform for the developed system. Although we were successful in developing such a system that will run efficiently on a UNIX-based work-station, current personal computer X server software are not capable of running the system efficiently.

Immortalization of normal liver functions in cell culture: rat hepatocyte-hepatoma cell hybrids expressing ornithine carbamoyltransferase activity.

Normal rat hepatocytes have been fused with highly differentiated rat hepatoma cells. Some of the hybrids express a physiologically significant level of activity of the urea cycle enzyme ornithine carbamoyltransferase (OCT), a liver-specific function not found in the hepatoma cells. These hybrids have 10% of the adult rat liver OCT specific activity, incorporate 3H-ornithine into protein arginine, and can be selectively grown in arginine-free medium supplemented with ornithine. Somatic cell hybridization of normal differentiated cells with highly differentiated neoplastic cells of the same tissue type may be useful as a general method for obtaining permanent cell lines with new tissue-specific phenotypes.

Electrocardiographic body surface potential maps of the QRS and T of normal young men. Qualitative description and selected quantifications.

A qualitative and quantitative analysis of the Body Surface Potential Maps (BSPM) of 40 young men, ages 19-41, is presented utilizing a 180 electrode system, with 135 anterior leads and 45 posterior leads. Evidence for epicardial right ventricular breakthrough was demonstrated in 36/40 at 27.9 +/- 6.8 ms, whereas our previous studies on normal children (average age 12.5 years) have demonstrated evidence for epicardial right ventricular breakthrough at 25.0 +/- 8.9 ms. The peak-to-peak magnitude at maximal potential (at 42.3 +/- 4.8 ms) was not significantly different from that of the children (4,430 +/- 1,165 microV), and the peak-to-peak magnitude of ST-T was virtually the same as that of the children (1,182 +/- 376.2 microV). The terminal activation pattern of late QRS on the body surface map appeared in the true posterior, anterior superior, posterior right superior and/or right anterior superior positions, in order of frequency. There were other regions appearing less frequently. In contrast, this pattern in children was seen only in the anterior superior, right anterior superior, posterior right superior, and true posterior in order of frequency. In 18/40, the body surface manifestation of repolarization was seen an average of 9.4 +/- 4.8 ms before the end of the QRS. A new pseudocolor display with 31 color levels representing body surface potentials allowed excellent resolution of isopotential detail.

An evaluation of pulsus alternans in closed-chest dogs.

Pulsus alternans is a condition in which the arterial pressure generated by the heart oscillates between two levels on a beat-to-beat basis. We evaluated the onset of pulsus alternans in chronically instrumented dogs subjected to tachycardia and inferior vena caval occlusion. During pulsus alternans, the left ventricular (LV) end-diastolic volume (EDV) was larger before the strong beats (28.7 +/- 5.3 vs. 25.9 +/- 4.5 ml, P less than 0.001 by paired t test), suggesting that the Frank-Starling mechanism participates in the alternating difference in end-systolic pressure. In addition, however, the ratio of pressure to volume at end systole was greater in the strong beats (2.01 +/- 0.36 vs. 1.46 +/- 0.45, P less than 0.005 by paired t test), a difference that cannot be explained by the Frank-Starling mechanism alone. This indicates that there is also a difference in end-systolic inotropic states between strong and weak beats. These changes occurred without significant alterations in beat-to-beat levels of coronary flow. The time constant of isovolumic pressure fall (T) was faster for the strong beats (37.5 +/- 4.2 vs 61.1 +/- 12.7 ms, P less than 0.002 by paired t test). The onset of oscillation in T preceded the onset of changes in LVEDV and LV systolic pressure in every case by an average of seven beats (range 3-11), suggesting that abnormalities of intracellular calcium handling led to the occurrence of pulsus alternans.(ABSTRACT TRUNCATED AT 250 WORDS)

An analysis of variability of left ventricular pressure decay.

This study evaluated whether the time course of left ventricular (LV) pressure decay is consistent from beat to beat in the normal heart under tightly controlled experimental conditions. We determined the variability of LV isovolumic relaxation and compared it with that of other hemodynamic parameters. Pressure decay was evaluated using a monoexponential time constant (T), a half-time (T1/2), and an average rate (Ravg) in nine chronically instrumented dogs. To eliminate physical factors that could lead to variability, the dogs were studied at paced heart rates after autonomic blockade and during apnea. At a heart rate of 160 beats/min the coefficient of variation (SD/mean, expressed as a percent) was higher for T (4.7%, P < 0.005), T1/2 (5.0%, P < 0.005), and Ravg (3.2%, P < 0.005) than for dP/dtmax (1.9%), as well as for end-diastolic volume (1.2%), end-systolic volume (1.2%), or end-systolic pressure (1.8%). Similar differences were present at 200 beats/min. Pressure decay was also assessed during major loading shifts induced by rapid caval occlusion. Surprisingly, comparison of first and last beats did not show significant differences for T or T1/2 but did for all standard hemodynamic parameters and for Ravg. While the best correlation with a relaxation parameter and hemodynamic parameters during changing loading conditions was for Ravg, the correlations were not consistent in every case. We conclude that LV pressure decay shows marked variability, unrelated to the algorithm used to assess it. Ravg, a model independent parameter, may be a useful way to quantify LV pressure fall.

Infection of an implantable cardioverter defibrillator: management without removal of the device in selected cases.

A case is presented in which an implantable cardioverter defibrillator (ICD) became infected in the abdominal wall pocket 5 weeks following implantation. There was no evidence clinically or by computed tomographic scan suggesting mediastinal extension of the infection. The infection was treated successfully by debriding the abdominal wall pocket followed by a combination of pocket irrigation with antibiotic solution, parenteral antibiotics, and long-term oral antibiotics. This conservative therapy was successful and avoided removal of the device.