Federated Learning for COVID-19 Detection: Promising Artificial Intelligence Research

Strategic Insights -

Researchers at the University of Medicine and Pharmacy Craiova have developed a federated learning framework using pre-trained deep learning models to detect COVID-19 from unsegmented chest CT images. This approach demonstrates promising results, with a centralized VGG-16 model achieving a training categorical accuracy of 93.90% and a validation accuracy of 79.00%. The proposed federated VGG-16 model attains a training categorical accuracy of 83.82% and a validation accuracy of 79.32%. These findings suggest that federated learning can effectively facilitate collaborative model development across institutions while preserving data privacy, offering a viable adjunct diagnostic tool to enhance COVID-19 detection and patient management.

Key Takeaways:

  • The researchers developed a federated learning framework to detect COVID-19 from unsegmented chest CT images using pre-trained deep learning models.
  • The framework demonstrated promising results with a centralized VGG-16 model achieving a training categorical accuracy of 93.90% and a validation accuracy of 79.00%.
  • The proposed federated VGG-16 model attained a training categorical accuracy of 83.82% and a validation accuracy of 79.32%.
  • The study underscores the importance of data privacy in collaborative model development, highlighting the potential of federated learning in facilitating joint research efforts while preserving sensitive information.
  • The research has implications for the development of adjunct diagnostic tools to enhance COVID-19 detection and patient management.
  • The dataset used in the study consisted of 2,230 axial chest CT images, categorized into three groups: COVID-19 (1,016 images), lung cancer and non-COVID-19 lung infections (610 images), and normal lung appearances (604 images).
  • The COVID-19 images were sourced from the University of Medicine and Pharmacy Craiova's picture archiving and communication system (PACS), as well as reputable public databases such as Radiopaedia, Radiology Assistant, Harvard Dataverse, and the COVID-19 common pneumonia chest CT dataset.

Statistics:

  • 2,230 axial chest CT images were used in the dataset, categorized into three groups: COVID-19 (1,016 images), lung cancer and non-COVID-19 lung infections (610 images), and normal lung appearances (604 images).
  • The centralized VGG-16 model achieved a training categorical accuracy of 93.90% and a validation accuracy of 79.00%.
  • The proposed federated VGG-16 model attained a training categorical accuracy of 83.82% and a validation accuracy of 79.32%.

Sources:

  • NewsRx. New Artificial Intelligence Research Reported from University of Medicine and Pharmacy Craiova (Federated Learning for COVID-19 Detection: Artificial Intelligence-Assisted Diagnosis from Unsegmented Chest Computed Tomography Scans). Medical Letter on the CDC & FDA. July 6, 2025; p 91.
  • Applied Medical Informatics. Federated Learning for COVID-19 Detection: Artificial Intelligence-Assisted Diagnosis from Unsegmented Chest Computed Tomography Scans. 2025,47(Suppl. 1). (Applied Medical Informatics - http://ami.info.umfcluj.ro)