Healthy brain patterns improve EEG depression detection accuracy
Normative Deviation Modeling for EEG-based Depression Detection
Machine Learning
Summary
Detecting depression using brain wave recordings from EEG is hard because people's brain signals vary a lot and come from different machines. The authors looked at using patterns from healthy people to spot differences that might show depression. They created two ways to measure these differences and combined them with deep learning to improve diagnosis. Their methods worked well when tested on different groups and EEG setups.
What this means in practice
- •For digital health platform developers: Incorporate healthy brain pattern deviation models to enhance EEG-based depression screening across diverse data sources without retraining.
- •For mental health screening teams: Use deviation-guided EEG analysis to improve accuracy of depression detection across different EEG devices and patient populations.
Authors
Wan Jiang, Jingqi Cheng, Hui Zhang, Xu Zhao, Xiaojing Chen, Shengeng Tang, Jingjing Wu, Richang Hong
Abstract
Web-enabled digital health services are expanding access to physiological monitoring and mental health screening. However, electroencephalography (EEG)-based major depressive disorder (MDD) detection remains challenging across heterogeneous data sources, where individual variability and acquisition differences may affect disease-related representations. In this work, we investigate the use of healthy normative references for EEG-based MDD detection. Specifically, structural and distributional reference models are independently trained using healthy control (HC) recordings to characterize departures from healthy EEG patterns. We develop two diagnostic approaches that incorporate normative deviations at different representation levels. Scalar-level Normative Deviation-guided Diagnosis (S-NDD) summarizes structural and distributional deviations into scalar scores and integrates them with deep EEG representations. Feature-level Normative Deviation-guided Diagnosis (F-NDD) retains dimension-wise residuals and combines them with deep EEG representations through attention-based aggregation and auxiliary supervision. We also employ target HC-assisted score calibration for cross-source evaluation without updating the trained models. Experiments on two public EEG datasets examine both approaches under subject-independent within-source and bidirectional cross-source settings. The results demonstrate the effectiveness of normative deviation-guided diagnosis across different evaluation settings.