RT info:eu-repo/semantics/article
T1 Applying Deep Learning on a Few EEG Electrodes during Resting State Reveals Depressive States: A Data Driven Study
A1 Jan, Damián
A1 Vega Rodríguez, Manuel de
A1 López Pigüi, Joana
A1 Padrón González, Iván
A2 Psicología Evolutiva y de la Educación
K1 depression
K1 long short-term memory
K1 convolutional neural network
K1 gated recurrent unit
K1 non-linear features
K1 diagnostic tool
AB The growing number of depressive people and the overload in primary care servicesmake it necessary to identify depressive states with easily accessible biomarkers such as mobileelectroencephalography (EEG). Some studies have addressed this issue by collecting and analyzingEEG resting state in a search of appropriate features and classification methods. Traditionally, EEGresting state classification methods for depression were mainly based on linear or a combinationof linear and non-linear features. We hypothesize that participants with ongoing depressive statesdiffer from controls in complex patterns of brain dynamics that can be captured in EEG resting statedata, using only nonlinear measures on a few electrodes, making it possible to develop cheap andwearable devices that could be even monitored through smartphones. To validate such a perspective,a resting-state EEG study was conducted with 50 participants, half with depressive state (DEP)and half controls (CTL). A data-driven approach was applied to select the most appropriate timewindow and electrodes for the EEG analyses, as suggested by Giacometti, as well as the most efficientnonlinear features and classifiers, to distinguish between CTL and DEP participants. Nonlinearfeatures showing temporo-spatial and spectral complexity were selected. The results confirmed thatcomputing nonlinear features from a few selected electrodes in a 15 s time window are sufficientto classify DEP and CTL participants accurately. Finally, after training and testing internally theclassifier, the trained machine was applied to EEG resting state data (CTL and DEP) from a publiclyavailable database, validating the capacity of generalization of the classifier with data from differentequipment, population, and environment obtaining an accuracy near 100%.
YR 2022
FD 2022
LK http://riull.ull.es/xmlui/handle/915/35605
UL http://riull.ull.es/xmlui/handle/915/35605
LA en
DS Repositorio institucional de la Universidad de La Laguna
RD 17-may-2024