This paper introduces HCC-STAR, a clinically aligned LLM for hepatocellular carcinoma that converts routine EMR narratives into risk-based staging, ranked guideline-consistent treatments with evidence rationales, and individualized survival estimates. The model was trained using about 30,000 SEER cases expanded into clinician-validated EMR-style narratives and evaluated on 6,668 patients from 12 hospitals in China. The abstract reports state-of-the-art treatment recommendation and risk-stratification performance against guidelines, GPT-5, Gemini-2.5 Pro, residents, and attending physicians. It also reports hypothetical median survival of 51 months under HCC-STAR recommendations versus 29 and 32 months for BCLC and CNLC strategies; this is not presented as prospective causal evidence.
No heat snapshots are available in the last 24 hours.