Application of Ensemble Learning for Respiratory Ailment Diagnosis: Case Studies on Biomedical and Chest X-ray Image Datasets

The rapid identification of respiratory ailments, such as lung cancer and COVID-19, is critical for timely intervention. Chest X-rays (CXR) serve as an accessible diagnostic tool; however, existing machine learning models often struggle with limited accuracy and sensitivity. This study proposes an ensemble learning-based approach for classifying respiratory ailments using both biomedical and image-based data. Three biomedical datasets and one CXR dataset are utilized as case studies. Histogram of Oriented Gradients (HOG) and Radiomics techniques are applied to extract features from CXR images, which are then processed using Principal Component Analysis (PCA) for dimensionality reduction. To enhance model performance, the Taguchi method is used to tune the parameters of multiple classifiers, including Convolutional Neural Networks (CNN), Support Vector Machine (SVM), Decision Tree (DT), Naïve Bayes (NB), K-Nearest Neighbors (KNN), and Tree Bagger (TB). The proposed ensemble learning approach outperforms individual classifiers by at least 10%, demonstrating significant improvements in accuracy, sensitivity, specificity, precision, recall, F-measure, and G-mean. Statistical tests, including the Wilcoxon Signed-Rank Test and ANOVA, are employed to determine the optimal train-test split and validate the efficiency of the applied methods. The results highlight the potential of ensemble learning in improving diagnostic accuracy for respiratory ailments.