Computer-Aided Diagnosis System On Normal Liver Diseases Based On Multi-phase CT Image

With the fast development of modern society, the medical community has observed an obvious hepatic morbidity. The liver cancer, as one of the diseases of highest incidence rates, is devouring thousands of lives, for reason that the corresponding pathological features of early cancer are relatively unobvious and the diagnosis often happens in the late stage. Nowadays, the clinic applies the paracentesis in confirming the liver cancer, which causes the unbearable psychological and physical suffering for patients. So it will be extremely helpful if the computer-aided diagnosis can be used in the medical treatment of hepatic diseases.A computer-aided automatic diagnostic system, which including the region of interest (ROI) extraction, feature extraction and selection and categorizer design, is presented. Since the similarity of different hepatic diseases in ordinary CT scan may cause reduction in precision rate, the proposed system uses multi-phase abdomen CT as input. Firstly, the lesion is captured as ROI from the abdomen CT using level set method and region growing method; then the feature vectors are extracted based on grey level histogram and gray level co-occurrence matrix and the timing characteristic based on multi-phase CT image, and these vectors are selected using principal component; at last, after dimension reduction optimization, the feature vectors are put into the categorizer module which uses the support vector machine as the algorithm. The categorizer is designed as a triple-decker binary classifier with the results of diagnostic rate of normal, abnormal, hepatic cysts, hepatic hemangioma, liver cancer and other possibility. The categorizer also offers the characteristic curve value of manipulator which can also be used as the indicator of the system performance. The stability and high accuracy rate of diagnosis have been proved by the experimental data. As the most important indicator of the system, the classification accuracy of normal and abnormal has achieved99.49%, which shows the reliability and effectiveness of the method in this paper.According to the feedback from the clinical, the proposed system has proved its value, and it could be further modified in liver segmentation and lesion extraction to improve the diagnosis performance.