Scholarlink

<p>The COVID-19 pandemic has increased the demand for rapid and accurate diagnostic methods worldwide. This study presents a machine learning-based system that can serve as an effective diagnostic tool for classifying chest X-ray images into three categories: standard, viral pneumonia, and COVID-19. It employs well-known feature extraction techniques, including the grey-level co-occurrence matrix (GLCM), the grey-level run length matrix (GLRLM), the grey-level size zone matrix (GLSZM), and other descriptors that capture geometrical and textural variations in the chest. To train and evaluate various machine learning classifiers, this work uses 10-fold cross-validation. The classifiers include support vector machines (SVM), random forests (RF), K-Nearest Neighbors (KNN), XGBoost, and stacking ensemble methods. The best-performing model, stacking, achieved an accuracy of 89.43%, a ROC-AUC of 97.72%, a Matthews Correlation Coefficient (MCC) of 84.15, and a Kappa score of 84.15, significantly outperforming other models. These metrics effectively balance sensitivity and specificity for COVID-19 detection, especially when working with imbalanced datasets. Results indicate that combining GLCM, GLRLM, and GLSZM feature extractors greatly enhances model performance. This diagnostic model is not only reliable, interpretable, and computationally efficient but also suitable for real-time implementation in resource-limited 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

Advanced feature extraction techniques for machine learning based respiratory illnesses detection in chest radiographs

  • 期刊论文

  • 发表日期：

    2025年10月9日

查看全文
  • Kapila Gurucharan,
  • Alluri Sreenivas

摘要

The COVID-19 pandemic has increased the demand for rapid and accurate diagnostic methods worldwide. This study presents a machine learning-based system that can serve as an effective diagnostic tool for classifying chest X-ray images into three categories: standard, viral pneumonia, and COVID-19. It employs well-known feature extraction techniques, including the grey-level co-occurrence matrix (GLCM), the grey-level run length matrix (GLRLM), the grey-level size zone matrix (GLSZM), and other descriptors that capture geometrical and textural variations in the chest. To train and evaluate various machine learning classifiers, this work uses 10-fold cross-validation. The classifiers include support vector machines (SVM), random forests (RF), K-Nearest Neighbors (KNN), XGBoost, and stacking ensemble methods. The best-performing model, stacking, achieved an accuracy of 89.43%, a ROC-AUC of 97.72%, a Matthews Correlation Coefficient (MCC) of 84.15, and a Kappa score of 84.15, significantly outperforming other models. These metrics effectively balance sensitivity and specificity for COVID-19 detection, especially when working with imbalanced datasets. Results indicate that combining GLCM, GLRLM, and GLSZM feature extractors greatly enhances model performance. This diagnostic model is not only reliable, interpretable, and computationally efficient but also suitable for real-time implementation in resource-limited healthcare settings.