An Expert System for Differential Diagnosis of Myocardial Infarction

Abstract

Automated early detection of myocardial infarction (MI) has been long studied for the purpose of saving human lives. In this paper, we propose a rule-based expert system to analyze a 12-lead electrocardiogram (ECG) for various types of MI. This system is developed by mapping clinical definitions of different types of MI and their differential diagnosis into corresponding algorithmic rule sets. Essential preprocessing steps such as baseline correction, removal of ectopic beats, and median filtering are carried out on recorded ECG. Techniques such as multistage polynomial correction and QRS subtraction are exploited to achieve reliable preprocessing. The processed ECG is then delineated using a time-domain differential-based search algorithm recently proposed by the team to obtain the relevant features and measures. These features and measures are further utilized by an if-then rule set to classify the ECG into various groups. The performance of the system when validated on sample MI database exhibited a sensitivity of 95.7% and specificity of 94.6%. Unlike many previous works, this reliable performance is achieved without the use of abstract classifiers or the need of prior training. Being based on medical definitions, the system is also easily comprehensible, modifiable, and compatible with manual diagnosis.