Work

Multi-stage deep transfer learning with style generative adversarial network (StyleGAN) for abnormality localization of chest radiographs

Year: 2026

Type: article

Abstract: The chest X-ray has been shown to play a significant part in the classification of chest diseases, including but not limited to covid19, pneumonia, lung disease, and others. There is potential for further development in this area of research, which focuses on the classification of chest radiographs through the application of image processing analysis. Despite the fact that COVID19 has become endemic, the research on chest X-ray analysis has the potential to offer up new opportunities for evaluating, forecasting, and perhaps preventing diseases of a more severe nature in the future. The previous approach mostly focused only on improving the feature extraction process with various machine learning approach without enhancing the dataset itself. We proposed using a multistage deep transfer learning technique that incorporates an updated DCGAN and Style GAN in order to improve the Chest Radiographs dataset of COVID-19 patients. This would increase the accuracy of the dataset. The execution of multi-stage transfer learning is included in the proposal. This will be accomplished by the utilization of Inception V3, Xception, and VGG19, in that order. In addition, the application of sophisticated deep convolutional neural networks (DCGAN) for the purpose of data enhancement is considered to be an additional component. The results of the categorization show that a high level of competence is there, as evidenced by the fact that the F1 score was 98.64%, the accuracy rate was 98.96%, the precision rate was 98.39%, and the recall rate was 98.64%. There is a possibility that future research may concentrate on improving the quality of datasets and studying ways that can facilitate transfer learning through the utilization of alternate combinations of learning models. Both of these key areas have the potential to offer opportunities for study pertaining to the future. The ultimate goal of this research is to create a society that is defined by the highest possible levels of health and well-being. Enhancing the healthcare system in accordance with Saudi Vision 2030 will be the means by which this objective will be accomplished.

Source: Array

DOI: 10.1016/j.array.2026.100836

Authors: Ahmad Hoirul Basori, Sami Alesawi, Andi Besse Firdausiah Mansur, Purbandini P, Anas W. Abulfaraj, Sharaf J. Malebary

Institutions: King Abdulaziz University, Universitas Airlangga

Language: English

Topic: COVID-19 diagnosis using AI

Subfield: Radiology, Nuclear Medicine and Imaging

Field: Medicine

Domain: Health Sciences

SDG: Good health and well-being

Open Access status: gold

Estimated APC paid: $1,680