基于机器学习构建动脉瘤性蛛网膜下腔出血患者早期并发急性呼吸窘迫综合征预测模型

打开文本图片集

Abstract： ObjectiveTo develop a machine learning-based predictive model for acute respiratory distress syndrome（ARDS）within 7 days in patients with aneurysmal subarachnoid hemorrhage（ aSAH） based on common clinical indicators and scoring systems. MethodsThe data patients with aSAH were extracted from the MIMIC-IV 2.2 database，and key features were selected using the Boruta algorithm，Logistic regression，and Lasso regression. Eight machine learning algorithms were employed to construct predictive models，and model performance was evaluated using receiver operating characteristic （ROC） curves，area under the curve（AUC）， calibration curves， decision curve analysis（DCA）. The SHAP analysis was used to analyze model performance interpretation. ResultsThe study included 1 329 patients with aSAH，and 8 key features were identified，including pulse oxygen saturation，maximum blood glucose level，SOFA score，respiratory rate， Glasgow coma scale（GCS） score，serum creatinine，white blood cell count， and body temperature.After hyperparameter tuning the machine learning model，the XGBoost model achieved overall accuracies 80.9% on the training set and 77.1% on the validation set， with AUC O.883 and O.782，respectively，performing best among all evaluation metrics. SHAP analysis revealed that pulse oxygen saturation，maximum blood glucose，and SOFA score played crucial roles in the model's predictions. ConclusionsA predictive model for the early onset ARDS in aSAH patients was constructed based on the XGBoost algorithm. This model integrates multiple clinical variables and demonstrates excellent predictive performance，making it suitable for precise stratification early intervention in aSAH patients and optimization clinical treatment strategies.

Key words： aneurysmal subarachnoid hemorhage； acute respiratory distress syndrome;machine learning；prediction model； early diagnosis

动脉瘤性蛛网膜下腔出血（aneurysmalsubarachnoidhemorrhage，aSAH）是一种高致死性疾病，全球每年新发病例约6.1例/10万人，是威胁公共健康的重要疾病之一[1-2]。（剩余12496字）