A Federated Learning Framework for Mental Stress Assessment and Brain Source Localization Using EEG and fNIRS
Abstract
Abstract Mental stress influences how individuals think, perform tasks, and maintain long-term well-being, which makes its assessment a nontrivial problem in both clinical practice and everyday work environments. Although machine learning methods have been adopted in recent years to support stress analysis, many reported studies continue to rely on centralized training strategies. Such designs not only raise concerns about data privacy, but also tend to show unstable performance when models are applied across institutions with differing data characteristics. In response to these challenges, we develop a federated learning framework that integrates Graph Neural Networks (GNNs) with Physics-Informed Neural Networks (PINNs) to jointly address mental stress classification and brain source localization. In this framework, multimodal EEG and fNIRS recordings are obtained during carefully designed stress-induction tasks and are processed using artifact mitigation, normalization procedures, and event-aligned segmentation. The proposed GNN component explicitly models dynamic functional connectivity by representing brain activity as spatio-temporal graphs, enabling the learning of stress-related interaction patterns across sensors. In parallel, PINNs are employed to guide the source localization process by embedding head conductivity constraints and neuroanatomical priors, thereby improving spatial interpretability and localization accuracy. Model training is carried out in a…
Read the paper
Related studies
- Active Learning during Federated Learning for Object Detection
- Fairness-aware federated learning
- Decentralized Federated Learning
- Building a Personally Identifiable Information Recognizer in a Privacy Preserved Manner Using Automated Annotation and Federated Learning
- Preservation of Patient Level Privacy: Federated Classification and Calibration Models
- Efficient explainable federated learning
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.