1969 Volume 33 Issue 1 Pages 51-58
For the purpose of comparison in the diagnostic accuracy between the decision tree method and the joint probability method, they were applied to the diagnosis of the orthogonal electrocardiograms obtained from 101 patients with myocardial infarction (Infarction) and 601 patients without infarction (Non-infarction). As significant parameters, 14 measurements, including the duration and amplitude of Q wave, the deviation of S-T segment and the maximal T amplitude, were selected. A small-sized digital computer (NEAC-2203) was used for this study. When the correct diagnosis was made over 80 per cent in Infarction with the decision tree method, it was unavoidable for the rate of "false positive" diagnosis in Non-infarction to exceed 10 per cent. The accuracy of pattern classification in Infarction was 95.0 per cent with the joint probability method, and frequency of "false positive" diagnosis in Non-infarction was as low as 4.8 per cent. Thus, the better results in discrimination were obtained with joint probability method than with the logical decision tree method. These results indicate that the degree of diagnostic accuracy may be largely dependent on the classification method used and the joint probability method would be the better categorizer than the logical decision tree method in the automatic electrocardiographic interpretation.