Tuberculosis Detection in Chest X-Rays Using Hybrid Deep Learning Models

Tuberculosis remains a big global health challenge, especially in areas with a lack of high-tech medical diagnostics. This study proposes a hybrid deep learning model for TB detection using chest X-ray images and associated radiological reports via a built model that combines CNNs, ANNs, and RNNs. It leverages strengths of CNNs in extracting spatial features from images, ANNs for the representation of complex relationships, and RNNs for processing sequential text data from reports. We use a diverse dataset comprising 2,800 TB-positive X-rays, 3,500 normal ones, and 700 publicly available ones to ensure robustness and generalization. The performance for the proposed model reached 97 precent with a precision of 0.85, recall of 0.90, and F1-score of 0.88 in the most challenging cases. The authors have also used the Grad-CAM technique to visualize important areas on the X-rays responsible for the TB diagnosis to enhance interpretability. Because of the model's capability of integrating image and textual data, it can detect TB very early with high precision, hence making it a good tool to deploy, particularly in settings where resources are limited and timely intervention is very important.