Automatic learning of rules: A practical example of using artificial intelligence to improve computer-based detection of myocardial infarction and left ventricular hypertrophy in the 12-lead ECG

Abstract

The authors developed a computer program that detects myocardial infarction (MI) and left ventricular hypertrophy (LVH) in two steps: (1) by extracting parameter values from a 10-second, 12-lead electrocardiogram, and (2) by classifying the extracted parameter values with rule sets. Every disease has its dedicated set of rules. Hence, there are separate rule sets for anterior MI, inferior MI, and LVH. If at least one rule is satisfied, the disease is said to be detected. The computer program automatically develops these rule sets. A database (learning set) of healthy subjects and patients with MI, LVH, and mixed MI + LVH was used. After defining the rule type, initial limits, and expected quality of the rules (positive predictive value, minimum number of patients), the program creates a set of rules by varying the limits. The general rule type is defined as: disease = lim 11 < p 1 ≤ lim 1u and lim 2l < p 2 ≤ lim 2u and … lim nl < p n ≤ lim nu. When defining the rule types, only the parameters (p 1 … p n) that are known as clinical electrocardiographic criteria (amplitudes [mV] of Q, R, and T waves and ST-segment; duration [ms] of Q wave; frontal angle [degrees]) were used. This allowed for submitting the learned rule sets to an independent investigator for medical verification. It also allowed the creation of explanatory texts with the rules. These advantages are not offered by the neurons of a neural network. The learned rules were checked against a test set and the following results were obtained: MI: sensitivity 76.2%, positive predictive value 98.6%; LVH: sensitivity 72.3%, positive predictive value 90.9%. The specificity ratings for MI are better than 98%; for LVH, better than 90%.