Application of expert systems analysis to interpretation of fatal cases involving amitriptyline.

Abstract

Part I. Expert 4: The object of this study was to investigate the applicability of commercially available expert system shells to interpretation in forensic toxicology. Amitriptyline toxicology was selected as a pilot trial. Blood and tissue concentrations of amitriptyline and nortriptyline in fatal and nonfatal amitriptyline cases from the literature and from the Registry of Human Toxicology databank were entered into the expert system shell Expert 4 (Rivers, Elsevier). The statistical evaluation routines of the shell were used to search for patterns in the data. Successive changes in the data base were made to test for the influence of the data base on the conclusions. Finally the data base was refined, based on the evaluations, to strengthen the probabilities of the conclusions. The refined database was coupled with the Expert 4 inference engine to infer unknown parameters in the cases. The results of the expert system analysis were compared to known values and published expert opinions. The ratio of amitriptyline/nortriptyline and tissue levels of nortriptyline were found to be the most significant measures for interpretation of effect and time since ingestion. Part II. Computer Induction of Rules: Blood and tissue concentrations of amitriptyline and nortriptyline in fatal and nonfatal amitriptyline cases from the literature and from the Registry of Human Toxicology databank were entered into the expert system shell BEAGLE, (Forsyth, Machine Learning Research, Ltd.). The automatic rule induction routines of the shell were used to search for patterns in the data. The program expressed these patterns as numerical predictions or Boolean logic rules. The results were compared to those obtained with the Expert 4 (Rivers, Elsevier) using the same case knowledge base.(ABSTRACT TRUNCATED AT 250 WORDS)