This paper proposes a new fractal feature vector based on M-band wavelet transform for classification of ultrasonic liver images-normal liver, cirrhosis, and hepatoma. Classifications for liver images have revealed that the fractal feature vector is trustworthy. A hierarchical classifier produces 97.27% correct classification for the distinction of normal and abnormal liver image and 93.6% correct classification for the distinction of cirrhosis and hepatoma liver image.
展开▼