A machine learning algorithm for optimizing treatment selection for patients with up to three hepatocellular carcinomas measuring ≤ 3 cm.

Abstract

4019 Background: Liver resection (LR) and radiofrequency ablation (RFA) are recommended for patients with early-stage hepatocellular carcinoma (HCC) with three nodules measuring ≤3 cm and preserved liver function. This study aimed to develop a predictive model to guide treatment decisions based on survival outcomes. Methods: This study included 18,958 patients with up to three HCCs measuring ≤3 cm from the nationwide survey of Japan. The Recurrent Deep Survival Machines (RDSM) model was employed for deep survival analysis. We employed 10-fold cross-validation, the concordance index (C-index), and overall survival (OS) to assess model performance. Survival curves were compared using the log-rank test. To identify potential confounding factors, 1:1 propensity score matching (PSM) was performed. Results: Patients undergoing LR demonstrated significantly longer OS than those receiving RFA (5-year survival rate 81.4% vs. 73.1%; P < 0.005). The trained RDSM model achieved a C-index of 0.68. In the deep learning (DL) model, patients undergoing recommended treatment demonstrated significantly longer survival than those who did not (5-year survival rate 81.2% vs. 73.9%; P < 0.005; PSM, 81.9% vs. 76.7%; P < 0.005). The DL modeling recommended LR in 6,966 (84.5%) patients undergoing RFA, especially those showing typical imaging patterns (early enhancement and washout in the computed tomography images [85.9% vs. 61.7% and 84.1 vs.74.3%, respectively]). Conclusions: DL modeling effectively helped treatment allocation for patients with up to three HCCs measuring ≤3 cm. Our study indicates the potential utilization of DL modeling in the treatment allocation of patients with up to three HCCs measuring ≤3 cm.