Tuberculosis Diagnosis in Chest X-Rays via Transfer Learning Techniques

Tuberculosis (TB) remains one of the world's major health problems. Detection with Chest X-rays is a mainstay tool in the early diagnosis of TB. Traditional methods used in the conventional detection of TB are very slow and susceptible to human errors. Thus, there is a need for an automated technique. The paper discusses the possibility of using advanced deep learning techniques, particularly techniques stemming from transfer learning, to enhance the detectability as well as efficiency of TB. Fine-tuned pre-trained convolutional neural networks, such as ResNet and VGG, on a dataset of chest X-ray images extracted high-level features relevant to the detection of TB. Superhuman performance was achieved with limited medical imaging data by transferring learned representations from large-scale image classification tasks. Experimental results illustrate high accuracy, precision, and recall rates, confirming that transfer learning is adequate for automating TB detection. This approach reduces the need for manual intervention and offers a scalable, cost-effective solution to enable rapid TB screening, particularly in low-resource settings. The novelty of the work comes from using transfer learning with fine-tuned pre-trained models (ResNet and VGG) in getting high generalization accuracy for tuberculosis detection from chest X-rays, even with limited data.