Zhang, Shenghuan

McCane, Brendan

Neo, Phoebe S-H

Shadli, Shabah M

McNaughton, Neil

2020

<p>Purpose: This study aims to identify EEG biomarkers that predict the level of depressive personality (where extreme scores indicate disorder), as opposed to the presence or absence of a depressive state or a depression diagnosis. Methods: Fourier features were extracted from 2-second epochs of resting state EEG and used by LSBoost to maximise the correlation with depressive trait tendencies (PID-5 depressivity index). Results: Our method accounted for 25.75% of the variance in PID-5 scores, albeit in females only. The recording channel C3 and frequencies in the gamma band were the most important contributors to the prediction. The findings are consistent with previous psychological studies and suggest that our method is a feasible strategy for developing quantitative EEG biomarkers for trait depressivity in a neuropsychologically interpretable form. We have also shown that there might be different markers for depressivity between males and females.</p>

International Joint Conference on Neural Networks, 2020 Conference Proceedings

2161-4407

2161-4393

https://hdl.handle.net/1959.11/59737

en

Institute of Electrical and Electronics Engineers (IEEE)

Trait depressivity prediction with EEG signals via LSBoost

Conference Publication

Publication