| Popis: |
Objective To explore the application of machine learning in scientific and rational blood preparation and predictive analysis for surgical blood usage before liver transplantation surgery. Methods Clinical basic information including gender, age, clinical diagnosis and surgical methods of 356 liver transplantation patients were collected. The duration (Time) and preoperative laboratory test results of hemoglobin (Hb), hematocrit (Hct), platelet count (Plt), prothrombin time (PT), activated partial thromboplastin time (APTT), fibrinogen (Fib), total bilirubin (TBIL), albumin (ALB), creatinine (Crea) and total protein (TP), as well as the amount of intraoperative blood transfusion were collected. A machine learning model capable of predicting the risk of massive blood transfusion during liver transplantation surgery was established by Python, and was evaluated to select the optimal predictive model. Results Among the 7 machine learning models constructed, the logistic regression model performed the best (AUROC: 0.90, F1 score: 0.82), with an accuracy of 79.44% and precision of 79.69%, followed by the random forest classifier (AUROC: 0.87, F1 score: 0.83), with an accuracy of 79.44% and precision of 77.94%. Conclusion Establishing a machine learning prediction model by Python is of significant clinical importance for scientific blood preparation, predicting the risk of massive blood transfusion and ensuring the safety of blood use in liver transplantation surgery. |
