心X線陰影の計量診断

Abstract

Although each heart disease has its own characteristic configuration of cardiac silhouette in the chest X-ray film, its evaluation is made only empirically at present. The purpose of this study is to develop a quantitative method for diagnosis of cardiac silhouette. Eighty-six cases were employed for the study: 20 with mitral valvular disease, 13 with aortic valvular disease, 17 with atrial septal defect, 10 with ventricular septal defect, 8 with myocardial infarction and 18 healthy controls without any abnormalities in the circulatory system.Chest X-ray films were taken in 6 projections with intervals of 30 in the standing position at the fixed phase of systole using ECG-trigger apparatus. Cardiac silhouettes were measured by AMDCOX (automatic measurement by densitometry-computer system of X-ray) developed in our depertment before.The vertical distance from the intersection of sternum and the lower margin of the third rib to the intersection of the right cardiac border and diaphragm was divided evenly into 5 horizontal sections, and the horizontal width of the heart in each section was measured.Examining the frequency distribution functions of these values or their combinations in each disease group, 14 parameters were selected as useful items. Using these items, discriminant functions for simultaneous differentiation of 6 disease groups were derived by means of “Theory of Quantification”.According to this theory, the solution is given by 5 latent roots in case of differentiation of 6 disease groups. Namely, each individual case is expressed by a combination of 5 scores. In order to make the judgement easier, however, 3 graphs, each consisted of a pair of the scores, were used for practical differentiation.The scores in each disease group showed a characteristic distribution on these graphs, indicating that the diseases can be differentiated by the boundary lines which were drawn to separate them. By use of these boundary lines, all cases of mitral valvular disease, ventricular septal defect and myocardial infarction were diagnosed correctly. Misdiagnosis was made only in each one case of aortic valvular disease, atrial septal defect and healthy controls. Thus, the high rate of correct diagnosis of 96.5% was obtained (83 out of 86 cases). Three misdiagnosed cases were plotted in the vicinity very close to the boundary lines. This means that the misdiagnosis could be avoided by careful judgement in such cases.These results indicate the practical usefulness of the automatic and quantitative differentiation of the cardiac silhouettes in clinical medicine