Objective: To illustrate the utility of a rule-based expert system in diagnosing hemoglobin disorders.
Design: A rule-based expert system was developed for diagnosing hemoglobin disorders. This expert system runs on IBM-compatible personal computers and uses a backward-chaining search strategy to draw conclusions. Laboratory data (ie, results of hemoglobin electrophoresis, quantitative measurements of hemoglobin F and hemoglobin A2 levels, and result of a sickle cell screen) are processed by the system using defined rules to obtain a set of differential diagnoses. Additional data, such as hematologic parameters, ethnicity of the patient, and the presence or absence of certain clinical signs and symptoms, aid in making a final diagnosis. The rules in the current version of this expert system include diagnostic criteria for 71 hemoglobin disorders.
Setting: Regional academic medical center.
Patients: We tested the system by using 58 survey sample cases offered by the College of American Pathologists during the period of January 1989 through December 1994.
Main outcome measure: The established diagnosis for a given case must be included in the list of differential diagnoses suggested by the expert system.
Results: The expert system included the actual diagnosis as one of the top four differential diagnoses in 90% of the cases, whereas all the laboratories participating in the survey included it in 84% (mean) of the cases.
Conclusion: We propose that this user-friendly expert system is a potential tool for computer-assisted diagnosis of hemoglobin disorders.