Early diagnosis of Tuberculosis (TB) remains constrained by the shortage of radiologists and limited medical facilities. This study presents a solution through the implementation of the MobileNetV2 model to classify chest X-ray images into Normal and TB categories, integrated into a Flask- and React.js-based web interface. To balance the dataset to 10,000 images, the preprocessing stage applies Contrast Limited Adaptive Histogram Equalization (CLAHE), data augmentation, and pixel normalization. The model architecture was developed via transfer learning by fine-tuning the last 50 layers, and was subsequently evaluated using 5-Fold Cross-Validation alongside 1,500 independent test data. Model evaluation yielded an accuracy of 99.93%, precision of 100%, recall of 99.87%, and an F1-score of 99.93%, with a single false negative case detected. Regarding the application, Alpha Testing confirmed full functional success (100%), while Beta Testing with radiographers demonstrated a 95% feasibility rating (categorized as Highly Suitable). These findings demonstrate the effectiveness of MobileNetV2 as a web-based decision-support instrument for early TB screening.
Copyrights © 2026