Statistical Diagnosis of Electrocardiogram by Theorem of Bayes

Publisher Website
Google Scholar
Abstract
Diagnosis of the disease was undertaken statistically by ecg findings with the aid of the theorem of Bayes, without referring to other clinical and laboratory data. The incidence of 40 ecg findings together with age and sex were investigated in 17 diseases or groups of diseases in which cardiac disturbances play a principal role. From the incidence the probability of diseases was calculated in 25 actual examples by electronic computer. In cases with myocardial infarction, the computed results coincided with clinical diagnosis with a probability of about 100%. Hypertension and coronary sclerosis were hardly differentiated by electrocardiographic findings, although most cases were answered more or less correctly. The probability of a correct answer was not high in valvular diseases. When the correct answer was not obtained, however, other kinds of valvular diseases were answered with relatively high probability and hypertension or coronary sclerosis were ruled out almost completely. Fairly excellent results were obtained in cases with chronic cor pulmonale and neurocirculatory asthenia. There was a case in which the computed result did not coincide with clinical diagnosis, but with autopsy findings. Two or 3 diseases were strongly suspected by calculation in not a few cases which have no characteristic electrocardiographic findings. This fact suggests that ecg contains still undisclosed information. The stronger hemodynamic influence on the ecg than considered hitherto was suggested. It is our impression that statistical diagnosis of ecg''s are superior to our diagnostic ability in several respects.