AI Model HCC-STAR Improves Liver Cancer Treatment Guidance

Hepatocellular carcinoma (HCC), a prevalent and deadly cancer, currently relies on broad guidelines and staging systems that often overlook individual patient variations and the rich context found in electronic medical records (EMRs). To address this, a new clinically aligned large language model, HCC-STAR (Hepatocellular Carcinoma Staging, Treatment And pRognosis), has been developed. This model is designed to interpret EMR narratives and simultaneously generate risk-score based staging, ranked guideline-consistent treatment options with evidence-based justifications, and individualized survival predictions. The development involved curating approximately 30,000 HCC cases from the SEER database and augmenting them into EMR-style narrative training data using a clinician-validated, prompt-based workflow. HCC-STAR employs a knowledge-aligned reasoning framework, optimized with a step-verifiable composite reward, moving beyond simple text memorization of clinical guidelines. In a multi-center study involving 6,668 patients across 12 hospitals in China, HCC-STAR demonstrated state-of-the-art performance in treatment recommendation and risk stratification, surpassing both traditional clinical guidelines and advanced models like GPT-5 and Gemini-2.5 Pro. Hypothetical survival analysis indicated a median survival of 51 months with HCC-STAR recommendations, compared to 29-32 months with existing guidelines. Hepatobiliary specialists rated HCC-STAR's reasoning as trustworthy, and the model helped physicians make faster, more accurate decisions, even outperforming resident and attending physicians in treatment accuracy when used as an assistant.