Related Experiment Video
Updated: Nov 11, 2025
08:05
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
14.4K
Federated deep learning for detecting COVID-19 lung abnormalities in CT: a privacy-preserving multinational
Qi Dou1, Tiffany Y So2, Meirui Jiang3
1Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong SAR, China. qdou@cse.cuhk.edu.hk.
NPJ Digital Medicine
|March 30, 2021
Summary
Federated learning enables privacy-preserving artificial intelligence (AI) for COVID-19 detection using chest CT scans. This method ensures robust AI models generalize across multinational datasets, crucial for pandemic response.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Data Privacy
Background:
- Medical training databases require data privacy mechanisms for robust machine learning.
- Chest CT interpretation is vital for COVID-19 assessment and management.
- Federated learning offers a solution for decentralized, privacy-preserving AI model development.
Purpose of the Study:
- To demonstrate the feasibility of federated learning for detecting COVID-19 related CT abnormalities.
- To validate the generalizability of an AI model on multinational datasets.
- To explore federated learning for privacy-preserving AI in COVID-19 medical image diagnosis.
Main Methods:
- Recruited 132 patients from seven multinational centers (Hong Kong, Mainland China, Germany).
- Employed federated learning algorithms for training and external validation.
- Conducted case studies on longitudinal scans for lesion burden estimation.
Main Results:
- Demonstrated the feasibility of federated learning for COVID-19 CT abnormality detection.
- Achieved good generalization capability on unseen multinational datasets.
- Showcased automated estimation of lesion burden in hospitalized patients.
Conclusions:
- Federated learning is an effective mechanism for developing clinically useful AI across institutions and countries during pandemics.
- Overcomes challenges of central data aggregation for sensitive medical information.
- Facilitates rapid development of AI tools for infectious disease diagnosis and management.
