Harnessing Convolutional Neural Networks for Tuberculosis Identification in Chest X-ray Images

Tuberculosis (TB) continues to be a significant international health issue, with millions of new cases being recorded yearly. Early detection is crucial for efficient management and control of the disease’s spread. In response to this, our research focuses on the creation and assessment of a Convolutional Neural Network (CNN) for automated TB classification employing chest X-ray images. Our dataset comprises 700 publicly accessible TB photographs and 3,500 normal images, coupled with an additional 2,800 TB images retrieved via the NIAID TB site. The recommended CNN underwent training and validation, exhibiting solid performance on both sets. Results demonstrated good precision (98.90%) and recall (97.83%) on the validation set, highlighting the model’s capacity to appropriately detect TB-positive cases. Subsequent testing on an independent set further verifies the model’s dependability, reaching an overall accuracy of 99.52%. These findings highlight the potential of deep learning in aiding healthcare workers in TB diagnosis, contributing to timely intervention and improved patient outcomes. Our research highlights the necessity of utilising machine learning for enhancing TB diagnosis, delivering a crucial tool for healthcare practitioners and public health activities worldwide.