Developing and Validating an AI Model to Predict Best Post-Surgery Treatments for Hepatocellular Carcinoma Using Imaging and Pathology Data

What this Trial Is About

Hepatocellular carcinoma HCC is a common and deadly cancer worldwide, with a high rate of tumor recurrence after curative surgery, especially in China. Current postoperative treatments like transarterial chemoembolization TACE and tyrosine kinase inhibitors TKIs have limitations such as low response rates and significant side effects. This study seeks to develop a precise artificial intelligence AI model that integrates preoperative imaging, postoperative digital pathology, clinical data, and pathological reports to better predict which patients will benefit from specific postoperative therapies. The study includes a large retrospective cohort of about 10,000 HCC patients who underwent surgical resection, collecting detailed imaging, pathology, and clinical data to build and train the AI predictive model. Then, a prospective observational phase will enroll 1,000 postoperative HCC patients from multiple centers. The AI model will generate treatment predictions based on patient data but will not influence clinical decisions. Patients will be grouped by whether the actual treatments they receive match the AI predictions. Long-term follow-up will assess tumor recurrence and survival outcomes to validate the models prediction accuracy and clinical usefulness. Participants will undergo routine clinical care, including preoperative contrast-enhanced CT or MRI scans and postoperative collection of H&E-stained tissue slides for digital analysis. Researchers will collect comprehensive clinical, pathological, and follow-up data. Outcomes measured include recurrence-free survival and recurrence rates up to 3 years after surgery, with overall survival tracked for up to 5 years. No interventions are imposed by the study, and treatment decisions remain with physicians. The study aims to improve postoperative management and individualize adjuvant therapy for HCC patients through AI-based predictions.

Construction and Clinical Validation of a Predictive Model for Postoperative Adjuvant Therapy in Hepatocellular Carcinoma Based on Whole-Slide Digital Pathological Images and Deep Learning