Insightful Tuberculosis Detection: Integrating CNNs with Explainable LIME Methodology

One of the most transmissible illnesses and one of the top 10 causes of mortality globally is tuberculosis (TB). The effective treatment and control of tuberculosis (TB), which still poses a major danger to global health, depends on an early and accurate diagnosis. Although the manual interpretation of chest X-ray (CXR) imaging is laborious and prone to error, it is essential for the diagnosis of tuberculosis (TB). This work uses current breakthroughs in deep learning to build an automated method for TB detection from CXR photos. After collecting and preprocessing a range of annotated CXR picture datasets, convolutional neural network (CNN) architectures are investigated for model generation. Pre-trained models are refined on the dataset using transfer learning approaches, which are also included to improve diversity and reduce class disparity. Conventional metrics are utilized to evaluate the trained model on an unbiased test set by comparison to baseline methods and expert radiologist interpretations.