RI心血管造影像からの左心室輪郭抽出

抄録

Computer image processing techniques are applied to the radioisotope (RI) angiocardiographic left ventricular images.
In clinical medicine, the shape and dimensions of the left ventricle have been shown to provide many types of important information for the evaluation of the cardiac functions. Various techniques for this measurement have been devised. Recently, the RI-angiocardiography has come to be widely used, because it is superior to conventional X-ray method in its non-invasiveness.
However, images obtained by this method are usually quite noisy and poor in quality. It will be quite valuable for clinical use to apply the computer image processing technique to this image and to facilitate the extraction of features such as the left ventricular volume.
In this paper, properties of the RI-angiocardiographic image and several problems for the computer processing are first described. Then, processing techniques for the left ventricular boundary extraction are discussed. Two methods are introduced. One employs a radial scan and an ideal density curve fitting technique, the other employs a boundary tracing algorithm using the nonlinear edge enhancement technique. These methods have proved to work fairly well to locate boundaries in such a noisy image as RI-angiocardiogram.