The Center for Biomedical Informatics
State University of Campinas, Brazil

Research Abstracts

The diagnostic proficiency of general practicioners concerning convulsive and psychomotor disorders of epileptic origin is usually poor, particularly in developing countries. Thus, in order to provide a decision and learning tool for these cases, we have developed a computer program to automatically evaluate and classify patients with potentially epileptic episodes, with basis on the New International Classification of Epilepsy, using heuristic techniques. It is a rule-based expert system for IBM-PC compatible microcomputers, which is comprised of two parts: 1) the program itself, which is a programmable inference engine, or "shell" system developed by one of the authors (L.U.M.F.) ; 2) a disk-resident knowledge base comprised of IF...THEN production rules. The expert system is capable of classifying any potentially epileptic crisis into 12 different classes, using ca. 60 input data (symptoms and clinical signs collected from the patient's history) and ca. 120 rules. The domain-specific knowledge was acquired by structured interview and tree-graphing techniques from an experienced physician, and refined after extensive practical experimentation with real cases. The program combines forward and backward-chaining techniques, first to present relevant questions about the patient to the attending physician; following this it tries to ascertain whether the episode is of epileptic origin. Later on, it proposes one or more feasible classifications for the episode, using uncertainty factors. Both the input data and the program conclusions may be printed in hardcopy form or recorded onto disk for later review and analysis. The system is being extensively tested both by internist and experienced physicians in the outpatient clinic of the University Hospital, in order to investigate its usefulness as a decision-aid in a general clinical practice environment. To extend the program's utility in a medical environment of a developing country, where technological resources are scarce, the knowlegde bases use only patient complaints and objective, clinically observed symptoms and signs (no EEG or specialized laboratory findings are necessary); a further refinement of the knowledge base is underway, in order to include this data, as well as some degree of causal inference.

Published in:

Return to Home Page

Return to Abstracts Index

Last Updated: March 2, 1996

renato@sabbatini.com