Scholarlink

<p>Stroke remains a leading cause of long-term disability and mortality worldwide. Despite advances in medical imaging, diagnostic tools such as magnetic resonance imaging (MRI) and computed tomography (CT) remain expensive and are often inaccessible in resource-constrained regions. The objective of this study is to evaluate the effectiveness of machine learning (ML) and deep learning (DL) models in predicting stroke outcomes using non-imaging clinical data. We investigated multiple models, including artificial neural networks (ANN), random forests (RF), extreme gradient boosting (XGBoost), support vector machines (SVM), and an ensemble learning approach. Among these, the ensemble model outperformed all others, achieving an accuracy of 97.15%, an F1 score of 97.15%, and an area under the ROC curve (AUC) of 99.39%. ANN, RF, and XGBoost also performed well, with ANN reaching 96.41% accuracy and 99.03% AUC, although its recall was lower compared to tree-based models. The ensemble model demonstrated the highest robustness, achieving the lowest mean squared error (MSE: 2.43), root mean squared error (RMSE: 15.58), and mean absolute error (MAE: 6.86). These findings underscore the potential of ML-based solutions in enhancing stroke diagnosis, providing scalable and cost-effective alternatives to traditional imaging, especially in low-resource healthcare settings.</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

Stroke Prediction Using Ensemble Machine and Deep Learning Models

  • 期刊论文

  • 发表日期：

    2025年10月28日

查看全文
  • Md. Samiul Alom,
  • Rajib Kumar Halder,
  • Obaidul Haque,
  • Saikat Kumar Gain,
  • Md.Alamin Talukder,
  • Fahim Shakil Tamim,
  • Nusrat Jahan,
  • Tamanna Tabassum,
  • Fatema Akter,
  • Faisal Imran,
  • Rakib Hossen

摘要

Stroke remains a leading cause of long-term disability and mortality worldwide. Despite advances in medical imaging, diagnostic tools such as magnetic resonance imaging (MRI) and computed tomography (CT) remain expensive and are often inaccessible in resource-constrained regions. The objective of this study is to evaluate the effectiveness of machine learning (ML) and deep learning (DL) models in predicting stroke outcomes using non-imaging clinical data. We investigated multiple models, including artificial neural networks (ANN), random forests (RF), extreme gradient boosting (XGBoost), support vector machines (SVM), and an ensemble learning approach. Among these, the ensemble model outperformed all others, achieving an accuracy of 97.15%, an F1 score of 97.15%, and an area under the ROC curve (AUC) of 99.39%. ANN, RF, and XGBoost also performed well, with ANN reaching 96.41% accuracy and 99.03% AUC, although its recall was lower compared to tree-based models. The ensemble model demonstrated the highest robustness, achieving the lowest mean squared error (MSE: 2.43), root mean squared error (RMSE: 15.58), and mean absolute error (MAE: 6.86). These findings underscore the potential of ML-based solutions in enhancing stroke diagnosis, providing scalable and cost-effective alternatives to traditional imaging, especially in low-resource healthcare settings.