Enhanced Tuberculosis Detection in Chest X-Rays Using CNN with Coordinate Attention Mechanism

Tuberculosis or TB is a major global health risk, ranking the highest causes of death world-wide, making early detection crucial. Conventional methods for diagnosing TB through chest X-rays frequently lead to high rates of misdiagnosis because the disease often presents with subtle and hard-to-detect signs. The present study explores an advanced method for detecting tuberculosis, using a Convolutional Neural Network with a Coordinate Attention Mechanism, by leveraging the AlexNet architecture and fine-tuning it with the proposed attention mechanism, the model is able to capture spatial information more effectively, leading to improved accuracy in detecting TB. Extensive evaluations performed on a publicly accessible dataset reveal that the proposed model significantly surpasses traditional CNN methodologies in performance metrics, offering a promising tool for automated TB diagnosis.