Manuscript received April 12, 2024; revised July 2, 2024; accepted August 2, 2024; published January 17, 2025

Abstract—Pneumonia and tuberculosis are crucial reasons for infection and mortality worldwide, making lung illnesses a first-rate international fitness problem. Appropriate and active prognosis of those ailments is important for a successful remedy and for the care of the affected person. Because there are numerous similarities between lung illnesses and versions inside an unmarried condition, it may be tough to as it should be diagnosing lung illnesses. In this project, we endorse a singular technique to lung da sickness class method that divides lung X-ray pictures into 3 groups: “No Findings,” “Pneumonia,” and “Tuberculosis.” It does this by combining the energy of Visual Geometry Group (VGG) and Spatial Transform Network (STN). To capitalize on those neural community architectures` complementing advantages, our hybrid technique combines both. From the lung X-ray pictures, the VGG extracts low-degree features, that are in the end processed with the aid of using the Spatial Transform Network to gain extra complex correlations and contextual data among those features. We implemented a dataset of 12,856 chest X-rays to refine our model. On this dataset, we attained an accuracy of 94.59% implementing the technique we proposed. This shows how well our approach works for correctly identifying lung conditions from X-ray scans.

Keywords—Visual Geometry Group (VGG), Spatial Transform Network (STN), X-ray, Tuberculosis, Pneumonia, lung disease classification

Cite: Prasanthi A. Yavanamandha, Srikiran B. Kavuri, Sai Moulika C. Bedadhala, Vaishnavi D. Chintaparti, and Sindhu E. Kakumani , "Enhanced Lung Disease Detection using VGG and Spatial Transform Network (STN)," Journal of Image and Graphics, Vol. 13, No. 1, pp. 24-33, 2025.

Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC-BY-4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.