Abstract
INTRODUCTION: The objective of this study was to analyze the blood transfusion factors of minimally invasive direct coronary artery bypass (MIDCAB) surgery using artificial intelligence. METHODS: A retrospective analysis was performed for patients undergoing MIDCAB operations and no heart-lung machine was used from January 2017 to September 2022 in our hospital. The influencing factors of blood transfusion were used to build the artificial intelligence model. Eighty percent of the database was used as the training set, and twenty percent database was used as the testing set. To predict whether to use red blood cells during operation, we compared 104 artificial intelligence models. We aimed to assess whether which factors influence allogeneic transfusion in MIDCAB operations. RESULTS: Of the 104 machine learning algorithms, the XGBoost model delivered the best performance, with an AUC of 0.726 in the testing set and an accuracy of 0.854 in the testing set. The artificial intelligence model showed preoperative hemoglobin less than 120 g/L, prothrombin time greater than 13.75, body mass index less than 22.7 kg/m2, coronary heart disease with additional comorbidities, a history of percutaneous coronary intervention, weight lower than 67 kg were the six major risk factors of allogeneic transfusion. CONCLUSION: The XGBoost model can predict transfusion or not transfusion in MIDCBA surgery with high accuracy. INTRODUCTION: The objective of this study was to analyze the blood transfusion factors of minimally invasive direct coronary artery bypass (MIDCAB) surgery using artificial intelligence. METHODS: A retrospective analysis was performed for patients undergoing MIDCAB operations and no heart-lung machine was used from January 2017 to September 2022 in our hospital. The influencing factors of blood transfusion were used to build the artificial intelligence model. Eighty percent of the database was used as the training set, and twenty percent database was used as the testing set. To predict whether to use red blood cells during operation, we compared 104 artificial intelligence models. We aimed to assess whether which factors influence allogeneic transfusion in MIDCAB operations. RESULTS: Of the 104 machine learning algorithms, the XGBoost model delivered the best performance, with an AUC of 0.726 in the testing set and an accuracy of 0.854 in the testing set. The artificial intelligence model showed preoperative hemoglobin less than 120 g/L, prothrombin time greater than 13.75, body mass index less than 22.7 kg/m2, coronary heart disease with additional comorbidities, a history of percutaneous coronary intervention, weight lower than 67 kg were the six major risk factors of allogeneic transfusion. CONCLUSION: The XGBoost model can predict transfusion or not transfusion in MIDCBA surgery with high accuracy.