Computer evaluation of graphical physiologic data for diagnosis of coronary heart disease

Assessment of clinical significance in these records is sharpened by discriminant analysis. The selection and conditioning of data, requires hand measurements at kilobitS;-"OrIDachirle~ement at nearly megabit quantities, since multiple channels are essential. Clinically significant features of the graphical pattern 6f the BCG were determined, over adequate age-matched populations of normals and of patients with ischemiClleart disease. This required repeated calculation of the discriminant function, including several subsidiary statistics and graphical print-outs. Optimal features of a single physiologic trace thus produce a separatrix or a severity scale for the single disease-entity. Repeating this for other physiological traces, yields a sequential approach to diagnosis, in contrast to the uni-tary or parallel approaches of the Bayesian (probability-density) methods that have been used.