HCC Risk Prediction Model Based on Multimodal Data in Chronic Hepatitis B

We propose to construct and validate an early HCC risk prediction model in a multicenter retrospective cohort of hepatitis B-related fibrosis/cirrhosis patients, using multimodal data encompassing longitudinal clinical data, serum glycomics profiles, and liver biopsy histopathological images. Approximately 2,000 participants (200 HCC and 1,800 non-HCC) will be included. Model performance will be assessed by AUROC, sensitivity, and specificity through training and validation procedures, with the goal of developing a scalable HCC risk prediction method and software tool for clinical application.

Inclusion criteria

• Aged ≥18 years, both sexes eligible.
• Diagnosed with hepatitis B-related liver fibrosis or cirrhosis based on clinical assessment or liver histopathology.
• Undergoing antiviral treatment.
• Follow-up duration between 0.5- 5 years.
• Availability of baseline data and at least two longitudinal follow-up visits.
• Clinical outcomes during the follow-up period are traceable.

Exclusion criteria

• Diagnosis of Hepatocellular carcinoma (HCC) or liver transplantation at baseline or within 6 months after baseline/enrollment.
• Coexisting other chronic liver diseases, such as genetic/metabolic liver diseases, drug-induced liver injury, or severe steatotic liver disease.
• Coexisting other malignancies (except those considered cured).