ARCHIVES

Tuberculosis (TB) remains one of the leading causes of infectious mortality worldwide, underscoring the need for reliable and early diagnostic methods. This study presents an automated diagnostic framework that integrates Conditional Generative Adversarial Networks (CGANs) with Deep Transfer Learning (DTL) models to enhance early TB detection from chest X-ray (CXR) images. The CGAN is employed to generate high-quality synthetic TB-positive images, effectively addressing dataset imbalance and improving model generalization. Three pre-trained Convolutional Neural Network (CNN) architectures—DenseNet121, VGG16, and MobileNetV3—were fine-tuned using both original and CGAN-augmented datasets. Experimental results demonstrate notable improvements in diagnostic performance, particularly in accuracy, sensitivity, and F1-score, following data augmentation. Among the models, DenseNet121 achieved the highest classification accuracy of 97.8%. The proposed framework highlights the efficacy of combining generative augmentation with transfer learning for robust and accurate TB diagnosis, especially in data-constrained healthcare environments.