the Department of Medical Ultrasonics, Third Affiliated Hospital of Sun Yat-Sen University, 600 Tianhe Road, Guangzhou, China.

Liver diseases are worldwide problems including diffuse hepatitis disease and liver space occupying lesions. Hepatitis B virus infection is the most important one among difuse hepatitis disease. There are 240 million people infected with HBV globally, and more than one-third of these patients (approximately 93 million) live in China. For them, a precise estimation of the degree of liver fibrosis is important for estimation of prognosis, surveillance, and treatment decisions in patients with HBV infection. Hepatocellular carcinoma is usually detected in liver space occupying lesions and leading to third common death of cancer. Hence, precise diagnosis between benign and malignant tumor is of great significance. Many guidelines recommend ablation as first line treatment to small hepatocellular carcinoma, especially to those less than 1mm.

Liver biopsy has been considered the gold standard for assessing liver diseases. However, it is limited because of sample errors, interobserver variability and many complications, such as pain, bleeding and even death. CT/MR are alternative ways with high cost and complications. Ultrasound was a first optional tool to diagnose liver disease. However, there are still some problems remained to solve:first, the diagnosis accuracy of liver fibrosis remained to be improved. Second,the diagnosis performances for distinguishing benign and malignant tumor were poor. Third, there were not objective way to assess progression after liver ablations.

Radiomcis refers to the extraction and analysis of large amounts of image features from medical images. Previous studies showed that it can improve diagnosis accuracy and evaluating disease progression. Hence,we tried to combined radiomics and ultrasound images together in order to improve diagnosis performances of liver fibrosis, benign and malignant tumor and progression after liver ablations.