Retrospective Study to Develop a Model Using Artificial Intelligence to Predict Hepatocellular Carcinoma Risk in Cirrhosis

National Institute Of Diabetes And Digestive And Kidney Diseases

Investigators

Linked publications & trials

Abstract

The protocol was approved by the Institutional Review Board in September 2022. We have recruited approximately 500 cirrhosis subjects (among 20 have hepatocellular carcinoma) from NIH into the project and approximately 226 cirrhosis patients with hepatocellular carcinoma from Georgetown University Medical Center. Our collaborators have obtained the radiologic images, including abdominal MRI and ultrasound from both NIH and Georgetown and have started the image processing to determine which images are best to use to identify the radiologic features. We are building the clinical database of 500 patients from NIH, which include demographic factors, and clinical predictors previously known to predict liver cancer risk. The goals would be to first validate known existing models that have been used to predict liver cancer risk in cirrhosis patients in our cohort of NIH patients. Prior existing clinical models use baseline clinical predictors to predict long term hepatocellular carcinoma risk. Our model will use the radiologic changes in attempts to inform short term hepatocellular carcinoma risk. In conjunction to our team's database building, our NCI collaborators will first train a model using artificial intelligence to first differentiate liver cancer from non-liver cancer such as a cirrhosis background. Subsequently, they will then train model to predict liver cancer risk and we will refine the model with additional clinical predictors.