P-L02 Machine learning to evaluate liver reserve function based on venous blood biochemistry

Abstract Background The accurate and comprehensive evaluation of liver reserve function is crucial for daily follow-up medical treatment patients with disease. However, existing techniques are too complex costly universal implementation.To develop a convenient, reliable method to evaluate based on eight biochemical indicators measured from venous blood. Methods Blood test results (albumin (Alb), total bilirubin (TBIL), prothrombin time (PT), international normalized ratio (INR), cholesterol (TC), cholinesterase (ChE), aspartate amino transferase (AST), alanine transaminase (ALT)) were collected retrospectively 660 treated at the first hospital Ianzhou University 2016 2018. As reference standard function, indocyanine green (ICG) clearance also same times. patient data processed analyzed construct machine learning model, eXtreme Gradient Boosting (XGBoost), generalized linear model (GLM) predict indicators. Results showed that predicted XGBoost values closely correlated actual ICG 15-minute retention rates (R = 0.969, R2 0.939), while GLM had relatively low correlation 0.566, 0.320). These findings indicate developed can be used comparable performance test. Furthermore, exhibited superior prediction compared GLM. Hence, using utilized related commonly clinically, easier obtain than measures. Conclusions by highly when test, demonstrating strong practical clinical value model.