Early Tuberculosis Detection using Advanced Feature Extraction Techniques Integrated with a PSO-GWO-Based Neural Network Classifier

Abstract

The global health challenge posed by Tuberculosis (TB) necessitates innovative solutions for early detection, especially in regions heavily burdened by the disease. This paper introduces a groundbreaking approach—an automated computer-aided diagnosis system aimed at reducing reliance on expert radiologists for early TB detection through chest X-ray images. The proposed technique leverages advanced feature extraction methods, including GLCM, HOG, and DWT-GIST Descriptor coefficients, and employs a PSO-GWO based Neural Network (NN) classifier. This integration of sophisticated techniques contributes to a comprehensive and accurate TB detection system. The evaluation results demonstrate an impressive accuracy of 97.12%, highlighting the potential of the proposed approach to significantly improve the efficiency and accessibility of TB diagnosis, particularly in resource-constrained settings where expert radiologists may be scarce. This innovative system represents a crucial step towards addressing the challenges associated with TB diagnosis, offering a promising solution for timely and accurate detection.