LEAF(Liver Tumor dEtection And classiFication AI)

This study plans to utilize multiphase contrast-enhanced and non-contrast CT（Computed Tomography） images from 10000 pathologically confirmed liver tumor patients at our hospital. An AI（artificial intelligence） model will be used to outline the 3D contours of liver masses, which will then be refined by radiologists and hepatobiliary-pancreatic surgeons to enhance model accuracy. By incorporating more imaging data, the model's recognition capabilities will be improved, laying the groundwork for prospective clinical trials and aiming to establish a superior AI model for early liver cancer screening based on CT imaging.

This research project intends to utilize multiphase contrast-enhanced and non-contrast CT images from 10000 patients with a full spectrum of liver tumors (such as HCC（hepatocellular carcinoma）, ICC（intrahepatic cholangiocarcinoma ）, META（Metastasis）, etc.), confirmed by the pathological gold standard at our hospital. Through a pre-established AI model, the 3D contours of various liver masses will be delineated. In collaboration with senior physicians from our hospital's radiology department and hepatobiliary pancreatic surgery department, the AI-drawn contours will be refined to obtain more accurate 3D mass models, thereby enhancing the validation efficacy of the model. By incorporating more radiological data, the precision of the model will be improved, boosting its recognition capabilities and laying a solid foundation for subsequent prospective clinical trials: the research will be conducted over a period of two months. For cases where the model indicates malignancy without clear evidence from medical history or other data, follow-up will be performed to confirm the true value through pathological results. for cases where the model indicates malignancy and CT report malignancy clinical Soc procedure will be followed, and all patients with none malignany reported by model or CT report will be follow up via telephone every three mouths. The ultimate goal is to establish a superior AI model for early screening of liver cancer based on CT imaging.

Inclusion Criteria: * From 2019 to 2030, our hospital has collected non-contrast and contrast-enhanced CT images from patients with a full spectrum of liver tumors (such as HCC, ICC, META, etc.), all confirmed by the pathological gold standard Exclusion Criteria: * Patients who have undergone upper abdominal surgery. Examples include post-ERCP (Endoscopic Retrograde Cholangiopancreatography) for the pancreas, post-external drainage surgery, esophageal surgery, and gastrectomy, among others. * Patients who have received systemic treatments such as chemotherapy or traditional Chinese medicine. Examples include chemotherapy for lymphoma, chemotherapy for leukemia, chemotherapy for lung cancer, and comprehensive treatment for liver cancer, etc. * Patients with poor-quality CT images. Examples include convolution artifacts caused by the inability to place hands on the sides of the body and respiratory artifacts due to poor breath-holding, etc.