Abstract
As a global and grievous mental disease, major depressive disorder (MDD) has received much attention. Accurate detection of MDD via physiological signals represents an urgent research topic. Here, a frequencydependent multilayer brain network, combined with deep convolutional neural network (CNN), is developed to detect the MDD. Multivariate pseudo Wigner distribution is firstly introduced to extract the time-frequency characteristics from the multi-channel EEG signals. Then multilayer brain network is constructed, with each layer corresponding to a specific frequency band. Such multilayer framework is in line with the nature of the workings of the brain, and can effectively characterize the brain state. Further, a multilayer deep CNN architecture is designed to study the brain network topology features, which is finally used to accurately detect MDD. The experimental results on a publicly available MDD dataset show that the proposed approach is able to detect MDD with state-of-the-art accuracy of 97.27%. Our approach, combining multilayer brain network and deep CNN, enriches the multivariate time series analysis theory and helps to better characterize and recognize the complex brain states.
Original language | English |
---|---|
Pages (from-to) | 667-677 |
Number of pages | 11 |
Journal | Nonlinear Dynamics |
Volume | 102 |
Early online date | 11 May 2020 |
DOIs | |
Publication status | Published - Oct 2020 |
Bibliographical note
Acknowledgements This work was supported in part by the National Natural Science Foundation of China under Grants Nos. 61922062 and 61873181.Keywords
- Electroencephalogram
- Major depressive disorder
- Complex network
- Convolutional neural network
- FUNCTIONAL CONNECTIVITY
- FRAMEWORK
- CLASSIFICATION
- MODEL
- EPIDEMIOLOGY
- FEATURES