<![CDATA[Tuberculosis (TB) is a major global health concern and remains one of the deadliest infectious diseases, particularly in developing countries. Early and accurate diagnosis is crucial to initiate timely treatment, prevent complications, and reduce transmission rates. Conventional diagnostic methods, such as sputum tests and laboratory cultures, are often time-consuming and require specialized resources. Therefore, there is a growing need for automated, efficient, and accurate computer-aided diagnosis (CAD) systems. This study proposes a lightweight Convolutional Neural Network (CNN) architecture to classify chest X-ray images into TB and normal categories. The model is trained using the publicly available Shenzhen chest X-ray dataset, with three training durations: 10, 25, and 50 epochs. Although the model trained for 25 epochs achieved a slightly higher accuracy (86.36%) compared to the 10 epochs model (85.61%), the latter is considered more optimal due to its better balance between performance and efficiency. Specifically, the 10 epochs model produced high precision (92.86%) and a competitive F1-score (84.27%) while requiring significantly less training time and computational resources. Moreover, it maintained stable validation performance without signs of overfitting. In contrast, models trained for longer durations showed diminishing returns or performance degradation, particularly at 50 epochs. These results indicate that a shorter training cycle, when coupled with appropriate architectural design and regularization, can yield a robust and efficient classification model. This approach is particularly beneficial for deployment in resource-constrained environments, where rapid and reliable TB screening using chest X-ray images is critically needed.]]>
