Adaptive Ultrasound Image Matching for Automatic Liver Disease Diagnosis Using Generalized Hough Transform

The relatively low quality of clinical untrasound images limits the success of early disease detection and diagnosis based on the images. Filtering speckle noise to enhance the quality of ultrasound images is thus very important to construct an automatic clinical diagnosis system. In this paper, we present an automatic liver cirrhosis diagnosis system using adaptive ultrasound image matching. Given an input image, a preprocessing module is first performed to filter speckle noise and detect the extremely stable edge pattern which is defined by a similarity function using the generalized Hough transform. Furthermore, a template database is constructed to facilitate the detection of region of interests and liver cirrhosis classification. As compared with past approaches, we span a new way to unify the processing steps in a clinical application using the generalized Hough transform for ultrasound images. Experimental results show the effectiveness of the proposed method.