Prediction of postoperative mortality risk in hepatocellular carcinoma based on biomarkers and neural network models

Hepatocellular carcinoma (HCC) is one of the most common malignant tumors worldwide, with high morbidity and mortality. Postoperative survival is influenced by various factors. Traditional prognostic assessment methods have limitations and often fail to fully reflect individual patient differences. This study aims to analyze physiological, biochemical, and clinical characteristics of HCC patients to explore their association with mortality risk and construct a neural network-based HCC mortality scoring system, offering a novel approach to personalized prognostic evaluation. This retrospective study included 327 HCC patients admitted to a tertiary hospital between June 2020 and June 2022. Univariate and multivariate analyses were conducted to evaluate the relationship between biomarkers (e.g., lymphocytes, neutrophils, alanine aminotransferase [ALT], aspartate aminotransferase [AST]) and clinical features (e.g., number of lesions, imaging stages, microvascular invasion) with mortality risk. A neural network model was then constructed to develop an HCC mortality scoring system, and its predictive performance was assessed. Both univariate and multivariate analyses identified significant predictors of postoperative mortality risk, including lymphocytes (HR = 1.06, P < .001), ALT (HR = 0.98, P = .005), AST (HR = 1.03, P = .003), and the number of lesions (≥3, HR = 11.87, P < .001). The neural network model based on these factors demonstrated good performance in predicting survival among non-deceased patients (precision 78%, recall 84%), but its performance for the mortality category was suboptimal. Biomarkers such as lymphocytes, ALT, and AST, as well as clinical features including the number of lesions and imaging stages, are significant predictors of postoperative mortality risk in HCC patients. This study is the first to develop an HCC mortality scoring system using a neural network model, providing scientific support for personalized prognostic evaluation and precise treatment of HCC.