Computed-aid radiographic measurements of cardiac silhouette in Poodle dogs with mitral valve insufficiency

Abstract

A narrow range of normal variations in cardiac size and shape on radiographs is strongly influenced by technical factors, positioning and thoracic breed conformation. Although these alterations in cardiac silhouette can challenge the subjective analysis of cardiac silhouette, thoracic radiography is until consiterated a significant tool for complete evaluation of heart and cardiopulmonary diseases. Several measuring methods, therefore, have been proposed in order to improve the ability to detect cardiac enlargement and to minimize the subjectivity of empirical assessment. Application of image processing has satisfied these requirements, since it transforms qualitative features from human visual perception into quantitative measurements which can reflect anatomic changes and displacement of structures on radiographic image. In this work, three cardiac measurement methods will be considered: VHS (Vertebral Heart Size), proposed by Buchanan and Bücheler (1995), cardiac inclination and normalized cardiac area. The aim of this study is to compare the accuracy of the decision rules, based on each methodology separately, to the ones obtained using the combinations of two measurement methods out of the three. Seventy sets of thoracic radiographs of Poodles will be allotted to two groups: images (I) from dogs with no clinical signs of cardiopulmonary disease (n=10), and (II) from patients with proven cardiac disease (n=30), confirmed by echocardiography to be caused by degenerative mitral valve disease. The measurements will be performed, in right lateral view with computing techniques, from computer graphics and image processing, by one experienced veterinary radiologist and two trainees. To establish the angle, the space between the long axis of heart (L) and a straight line segment over the sternum (from fourth sternebrae to xyphoid process) that join each other, will be measured in degree. To determine normalized cardiac area (v2), the contours of the heart and of a square, which base corresponds to one fifth of the distance between five vertebrae (from T4 to T8), will be drawn, and the ratio of both areas will be obtained. The interobserver measurements will be compared using simple linear regression analysis. (AU)