Explainable GLCM–Decision Tree Framework for Pulmonary Tuberculosis Classification from Chest X-Rays
Pulmonary tuberculosis (TB) remains a major global health challenge, while chest X-ray interpretation is still influenced by radiologist expertise and inter-observer variability. This study develops an explainable machine learning framework for pulmonary TB classification from chest X-ray images by integrating Gray-Level Co-occurrence Matrix (GLCM) texture features, statistical image features, and a Decision Tree classifier implemented in RapidMiner. The dataset comprised 1,000 chest X-ray images representing TB-positive and normal classes. Image preprocessing included resizing, grayscale conversion, and intensity normalization, followed by extraction of four GLCM features—contrast, energy, homogeneity, and correlation—and four statistical features comprising mean, variance, standard deviation, and entropy. Model performance was evaluated using stratified 10-fold cross-validation. The Decision Tree achieved an average accuracy of 85.4%, precision of 87.1%, recall of 83.6%, and F1-score of 85.3%. Beyond predictive performance, the model generated interpretable IF–THEN rules that provide transparent reasoning for classification outcomes. Analysis of false-negative cases further identified clinically important misclassification patterns that require additional diagnostic confirmation. The proposed framework demonstrates that interpretable machine learning can provide a practical balance between classification performance, transparency, computational efficiency, and reproducibility, supporting its potential use as a clinical decision-support tool for TB screening in resource-constrained healthcare environments.