Modelling TB Detection Techniques using Chest X-rays in Zambia

Abstract

This study introduces a groundbreaking TB detection system in Zambia, harnessing chest X-ray analysis via machine learning. Addressing the limitations of traditional TB detection methods, this research utilized a layered architecture for image processing and classification using a convolutional neural network (CNN). The system's design and database architecture are tailored for scalability and maintainability, with an interface crafted for healthcare professional usability. A comprehensive hybrid dataset of 1,200 X-ray images, including 500 indicative of TB, was employed for the CNN model, achieving an impressive accuracy rate of 99%. However, the 10th epoch displayed an unusual drop in accuracy and recall, highlighting potential overfitting issues. The study's approach to TB detection using machine learning represents a significant leap in healthcare technology, especially in resource-limited settings like Zambia. By integrating advanced algorithms with medical imaging, the research paves the way for a more efficient, accurate TB diagnosis, potentially revolutionizing public health management in developing countries. The ROC curve and confusion matrix further elucidated the model's capabilities, ensuring its readiness for clinical application. This novel approach to TB detection not only streamlines the diagnostic process but also contributes to the broader narrative of technological integration in healthcare, setting new standards for disease detection and management.