Decision of Cirrhosis Using Liver's Ultrasonic Images

Feature extraction techniques in the ultrasonic images of a liver were studied firstly. As an initial result, total 25 parameters relating to cirrhosis were extracted. They were obtained from the motion curve of the liver in an M-mode image, and from the texture using a B-mode image. Then the efficiency of every parameter was analyzed, and with the use of a feature fusion method a set of 20 useful features was picked out. With these chosen features, a noninvasive automatic decision system for cirrhosis was constructed based on a Fisher linear decision criterion. The clinical application for 43 cases showed that this quantitative system was effective in the diagnosis of cirrhosis.