New model predicts treatment response in liver cancer therapy

A deep learning-based clinical-radiomics model has shown strong predictive capability for durable clinical benefit (dcb) in patients with potentially convertible hepatocellular carcinoma receiving immune checkpoint inhibitors (ICIs). The model achieved an area under the curve (AUC) of 0.96 in training and 0.88 in testing, accurately stratifying survival risk and revealing significant disparities in overall survival. Additionally, the model’s scores correlated with immune-related factors, highlighting connections to mutations and immune mechanisms in effective treatment outcomes.