The Heart Disease Program is a diagnostic program to assist the physician with the differential diagnosis of patients with cardiac symptoms, focusing on hemodynamic dysfunction.

Over the past dozen years, we have been developing a computer system to act as an intellectual sounding board, assisting the physician in the task of differential diagnosis and anticipating the effects of therapy in the domain of cardiovascular disorders. To address these problems we have developed two significant methodologies for medical reasoning. For diagnosis we have responded to the challenges of this very rich domain with a diagnostic mechanism that combines probabilistic reasoning in a Bayesian network with the constraints imposed by the severities of the states and the temporal relations of causality. This allows the Heart Disease Program (HDP) to generate differential diagnoses that are consistent with respect to the known conditions of causality in the medical domain. The hypotheses that make up the differential are causal networks representing the likely mechanisms causing and complicating the hemodynamic dysfunctions at a clinical level of detail. For predicting the effects of therapy we have developed a mechanism that uses equations for the hemodynamic relationships and a signal flow technique to calculate the likely quantitative steady-state change for all parameters given changes in therapies (or other parameter changes). This mechanism effectively captures the hemodynamic effects of the therapies on which it has been tested for a variety of pathophysiologic conditions.

There is a slide presentation about the program. There is also a slide presentation about the Web interface and one about temporal data.

The following are examples of the output of the program:

• Patient with syncope
• Patient with progressive dyspnea
• Patient with angina
• Patient with angina and dyspnea
• Patient with angina at rest

The program is experimental and is not to be used to diagnose patients. While we are constantly improving it, it is known to occasionally produce erroneous case analyses. The program is intended as a demonstration of the type of computer assistance that will be possible. We are making it available to determine how this functionality can best serve the physician. If you have appropriate medical training and know your way around murmurs, clicks, and rubs, you can try the program out. Heart Disease Program

The following are some of the publications describing the program:

• W. J. Long, S. Naimi, M. G. Criscitiello, G. Larsen ``Differential Diagnosis Generation from a Causal Network with Probabilities,'' Computers in Cardiology Conference, 1988.
  html

• William Long, ``Medical Diagnosis Using a Probabilistic Causal Network,'' Applied Artificial Intelligence, 3: 367-383, 1989.
  html or postscript

• William J. Long, ``Flexible Reasoning about Patient Management using Multiple Models,'' Artificial Intelligence in Medicine, 3: 3-20, 1991.
  html or postscript

• William J. Long, Shapur Naimi and M. G. Criscitiello, ``Development of a Knowledge Base for Diagnostic Reasoning in Cardiology,'' Computers in Biomedical Research, 25: 292-311, 1992.
  html or postscript

• William J. Long, Shapur Naimi and M. G. Criscitiello, ``Evaluation of a New Method for Cardiovascular Reasoning,'' Journal of the American Medical Informatics Association,1: 127-141, 1994.
  html or postscript

• William J. Long, ``Temporal Reasoning for Diagnosis in a Causal Probabilistic Knowledge Base,'' Artificial Intelligence in Medicine, 8: 193-215, 1996.
  html or postscript