Scholarlink

<p>Type 2 myocardial infarction (T2MI), distinguished from Type 1 myocardial infarction (T1MI) by oxygen supply - demand mismatch, has unique features. Acute kidney injury (AKI) following MI leads to severe consequences. Existing research mostly centers on T1MI, leaving a gap in T2MI related AKI studies. To address this, our research aims to explore AKI risk factors in T2MI patients and leverage machine learning algorithms to develop a model for accurate early prediction of AKI risk in this patient group. This retrospective study utilized the MIMIC-IV database (2008–2022) to analyze T2MI patients in critical care. The dataset was split 70:30 for model development. 12 machine learning algorithms underwent Boruta-based feature selection and hyperparameter optimization. All 12 machine learning algorithms were trained independently (i.e., no integration into a SuperLearner or other ensemble learning frameworks was performed). Model performance was assessed via AUROC, with SHAP analysis interpreting predictions, followed by web deployment for clinical use. Among 1,378 critically ill patients, 60.5% developed AKI post-ICU admission. Eleven variables were selected for machine learning modeling. XGBoost demonstrated superior predictive performance (test AUROC: 0.82, 95% CI 0.77–0.86). SHAP analysis identified mechanical ventilation as the strongest predictor, followed by minimum mean arterial pressure, maximum heart rate, maximum aspartate aminotransferase, and minimum white blood cell count. An interactive web tool (<a href="https://qhdpanguo.shinyapps.io/2MI-AKI/">https://qhdpanguo.shinyapps.io/2MI-AKI/</a>) was developed for clinical application.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning-based risk prediction model development for acute kidney injury in type 2 myocardial infarction patients

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月27日

查看全文
  • Pan Guo,
  • Lijing Xue,
  • Fang Tao,
  • Hongmei Yang,
  • Wenguang Wang,
  • Lixiang Ma

摘要

Type 2 myocardial infarction (T2MI), distinguished from Type 1 myocardial infarction (T1MI) by oxygen supply - demand mismatch, has unique features. Acute kidney injury (AKI) following MI leads to severe consequences. Existing research mostly centers on T1MI, leaving a gap in T2MI related AKI studies. To address this, our research aims to explore AKI risk factors in T2MI patients and leverage machine learning algorithms to develop a model for accurate early prediction of AKI risk in this patient group. This retrospective study utilized the MIMIC-IV database (2008–2022) to analyze T2MI patients in critical care. The dataset was split 70:30 for model development. 12 machine learning algorithms underwent Boruta-based feature selection and hyperparameter optimization. All 12 machine learning algorithms were trained independently (i.e., no integration into a SuperLearner or other ensemble learning frameworks was performed). Model performance was assessed via AUROC, with SHAP analysis interpreting predictions, followed by web deployment for clinical use. Among 1,378 critically ill patients, 60.5% developed AKI post-ICU admission. Eleven variables were selected for machine learning modeling. XGBoost demonstrated superior predictive performance (test AUROC: 0.82, 95% CI 0.77–0.86). SHAP analysis identified mechanical ventilation as the strongest predictor, followed by minimum mean arterial pressure, maximum heart rate, maximum aspartate aminotransferase, and minimum white blood cell count. An interactive web tool (https://qhdpanguo.shinyapps.io/2MI-AKI/) was developed for clinical application.