An Improvement of Liver Cirrhosis Classification Using Ultrasonic Images

概要

論文の詳細を見る
Ultrasonic images have been widely used for diagnosis of liver cirrhosis. In practice, there are significant differences in the result of a diagnosis among individuals for physician's experiences. Therefore, the objective and quantitative method for diagnosis is needed. The quality of M-mode image is in general low. Thus, it is very difficult to diagnose liver cirrhosis with M-mode images. In cirrhosis classification using the ultrasonic M-mode image, Zhou's method was proposed. In this paper, compared with the Zhou's method, we propose a method using weighted autocorrelation coefficient to extract an abdominal aorta wall. Furthermore, we propose the use of a bilateral filter to improve the quality of the M-mode images. The bilateral filter is expected to be effective one of smoothing filters preserving an edge to improve image quality in an ultrasonic image. The experimental results show that the weighted autocorrelation coefficient approach leads to improvement in the performance of the abdominal aorta wall extraction and moreover the bilateral filter is effective in both the abdominal aorta wall extraction and cirrhosis classification.