Scholarlink

Purpose <p>Coronary artery disease (CAD) remains a major global health challenge due to its high mortality rate. This study aims to enhance the reliability of CAD diagnosis by integrating conformal prediction (CP) with the non-dominated sorting genetic algorithm II (NSGA-II) by delivering a range of potential outcomes rather than a single deterministic result to reduce the risk of misdiagnosis by highlighting cases that may require further testing or intervention.</p> Methods <p>The proposed methodology employs NSGA-II to optimize two conflicting objectives: maximizing CP coverage and minimizing the size of the prediction sets. The framework is applied to the extended Z-Alizadeh Sani dataset (303 patients, 55 features) and utilizes the Taguchi method for systematic parameter optimization. Three machine learning models,&#xa0;extreme gradient boosting (XGBoost), <i>k</i>-nearest neighbors (KNN), and support vector machine (SVM),&#xa0;are evaluated at varying confidence levels to assess the diagnostic performance of the framework.</p> Results and conclusion <p>Experimental findings demonstrate that the NSGA-CP framework provides a robust and interpretable approach for CAD diagnosis, offering improved uncertainty quantification and reduced computational costs. This approach not only reduces the number of selected features but also significantly enhances diagnostic performance for detecting CAD in the left anterior descending, left circumflex, and right coronary arteries. The optimal Pareto solutions achieved a balanced trade-off between high CP coverage and minimal prediction set size.</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

Coronary artery disease diagnosis by integrating conformal prediction with a multi-objective evolutionary algorithm

  • 期刊论文

  • 发表日期：

    2025年7月22日

查看全文
  • Amir Mohammad Afzali,
  • Sarina Maleki,
  • Reza Tavakkoli-Moghaddam,
  • Ali Bozorgi-Amiri,
  • Keivan Tafakkori

摘要

Purpose

Coronary artery disease (CAD) remains a major global health challenge due to its high mortality rate. This study aims to enhance the reliability of CAD diagnosis by integrating conformal prediction (CP) with the non-dominated sorting genetic algorithm II (NSGA-II) by delivering a range of potential outcomes rather than a single deterministic result to reduce the risk of misdiagnosis by highlighting cases that may require further testing or intervention.

Methods

The proposed methodology employs NSGA-II to optimize two conflicting objectives: maximizing CP coverage and minimizing the size of the prediction sets. The framework is applied to the extended Z-Alizadeh Sani dataset (303 patients, 55 features) and utilizes the Taguchi method for systematic parameter optimization. Three machine learning models, extreme gradient boosting (XGBoost), k-nearest neighbors (KNN), and support vector machine (SVM), are evaluated at varying confidence levels to assess the diagnostic performance of the framework.

Results and conclusion

Experimental findings demonstrate that the NSGA-CP framework provides a robust and interpretable approach for CAD diagnosis, offering improved uncertainty quantification and reduced computational costs. This approach not only reduces the number of selected features but also significantly enhances diagnostic performance for detecting CAD in the left anterior descending, left circumflex, and right coronary arteries. The optimal Pareto solutions achieved a balanced trade-off between high CP coverage and minimal prediction set size.