LUNGCARE AI: ENHANCING TUBERCULOSIS DETECTION WITH MACHINE LEARNING

Abstract: Lung diseases such as Tuberculosis (TB) and Pneumonia pose major health challenges, particularly in countries like India. Chest X-rays are widely used for diagnosis, but manual interpretation can be time-consuming and inconsistent. This study presents a hybrid ensemble model combining InceptionV3 and VGG16 Convolutional Neural Networks (CNNs) for classifying TB, Pneumonia, and Normal lung conditions. The dataset undergoes advanced preprocessing and augmentation, followed by an 80:10:10 train-validation-test split. Using a majority voting strategy, the ensemble achieves around 95% overall accuracy and a 0.996 ROC AUC score. The model demonstrates strong potential for scalable, automated lung disease diagnosis in real-world clinical settings.

Keyword: Tuberculosis, Deep learning, Machine learning, Early detection.