Components of left ventricular dysfunction discriminated noninvasively beat by beat

Disturbed left ventricular contractility and synchrony were studied noninvasively beat by beat with simultaneous systolic time intervals and acceleration ballistocardiograms during quiet respiration. Thirty-six measures per beat, reduced to 7 by appropriate statistical methods, produced a linear discriminant function, based on a split training sample of 15 normal men (258 beats) and 9 men (149 beats) with proved myocardial infarctions. Validation tests on an independent set classified 89% of the beats from 12 tested normal subjects as "normal" and 96% of the beats from 6 tested men who had had infarction as "coronary-like," a physiologic, not etiologic, classification. In 14 hypertensive patients with no history or findings of coronary artery disease, 8 had 72 to 100% coronary-like beats per subject, suggesting unrecognized coronary artery disease, the cumulative effect of hypertension, or both. This method may help identify hypertensive patients at greater risk. Factor analysis found 4 discrete groups of the 7 discriminant variables that function together. Despite beat-to-beat variations due to respiration, myocardial dysfunction (when present) and individual differences, dysfunctional beats can be identified by these methods. They may also be applied to other multivariate data to extract the maximal information despite complex interactions.