P3DL: A Privacy Preserving Personalized Distributed Learning Framework for EEG-Based Cognitive State Identification.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Healthcare Technology

Background:

• Electroencephalography (EEG) is crucial for identifying cognitive decline in the elderly.
• EEG data contains sensitive personal information, posing privacy risks.
• Current methods prioritize accuracy over EEG data privacy.

Purpose of the Study:

• To develop a privacy-preserving personalized distributed learning framework (P3DL) for EEG-based cognitive state identification.
• To enhance both the accuracy and privacy of cognitive assessment in the elderly.

Main Methods:

• Proposed a privacy-preserving personalized distributed learning framework (P3DL) with clients and a central server.
• Implemented a federated dynamic update strategy (FedDBS) for model optimization.
• Introduced a novel loss function, extreme error Loss (E2Loss), to improve identification and misdiagnosis assessment.

Main Results:

• P3DL demonstrated an average increase in F2Score of 5.58% and 3.31% on clinical and public datasets, respectively.
• Accuracy improved by 1.78% and 2.46% on the respective datasets.
• Framework scalability was confirmed in emotion recognition tasks.

Conclusions:

• The P3DL framework effectively enhances cognitive state identification accuracy.
• P3DL ensures the privacy protection of sensitive EEG data.
• This work opens new avenues for secure and reliable healthcare applications using EEG.