FCEEG: federated learning-based seizure diagnosis through electroencephalogram (EEG) analysis
Electroencephalography (EEG) signals are crucial for seizure diagnosis. The data provides detailed insights into brain activity which aids in epilepsy management. Artificial intelligence (AI) and deep learning are widely employed in the analysis of EEG signals to achieve promising classification per...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Taylor & Francis Group
2025-12-01
|
| Series: | Cogent Engineering |
| Subjects: | |
| Online Access: | https://www.tandfonline.com/doi/10.1080/23311916.2025.2547636 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Electroencephalography (EEG) signals are crucial for seizure diagnosis. The data provides detailed insights into brain activity which aids in epilepsy management. Artificial intelligence (AI) and deep learning are widely employed in the analysis of EEG signals to achieve promising classification performance. However, these AI models require centralized data processing, thereby raising privacy concerns. Thus, we propose FCEEG, a convolutional-based deep learning with federated learning (FL) to diagnose seizures with EEG signals while preserving data privacy. Specifically, EEG data are learned and analyzed using convolutional neural networks (CNNs) on local clients without the need to transmit the clients’ raw EEG data to the central server. The decentralized process ensures the confidentiality and integrity of these sensitive health records. This balances data privacy with a promising performance. Additionally, this research involves experimenting with the best aggregation methods for EEG signals in federated learning. The empirical results demonstrate that our proposed framework FCEEG with Federated Proximal (FedProx) aggregation method can effectively utilize diverse local EEG data from local clients to perform reliable seizure detection with a promising performance with an accuracy of 87.66%, precision of 99.95%, specificity of 99.96%, recall rate of 75.86% and F1-score of 86.25%. |
|---|---|
| ISSN: | 2331-1916 |