Discriminant analysis of body surface potential maps for classification of non-Q wave infarction

Multivariate analysis was performed on 120 lead data from 159 normals (N), 233 patients with myocardial infarction (MI), and 116 patients with left ventricular hypertrophy (LVH). Optimal lead-ins (n=5) and features were identified for the multigroup model. The discriminant functions combined measurements from the P, QRS, and ST-T waveforms and included the duration of the P wave. The diagnostic performance achieved with this model averaged 89.2%. The multiple bigroup model required 7 leads, 6 of which were identical or closely related to those selected in the multigroup model. The diagnostic performance averaged 91.3%. The most striking difference between the schemes was observed when the discriminant functions were tested on patients from the non-Q-wave-MI group producing correct classification rates of 76% and 88%, respectively