SparNet: A Convolutional Neural Network for EEG Space-Frequency Feature Learning and Depression Discrimination

Abstract

Depression affects many people around the world today and is considered a global problem. Electroencephalogram (EEG) measurement is an appropriate way to understand the underlying mechanisms of major depressive disorder (MDD) to distinguish depression from normal control. With the development of deep learning methods, many researchers have adopted deep learning models to improve the classification accuracy of depression recognition. However, there are few studies on designing convolution filters for spatial and frequency domain feature learning in different brain regions. In this study, SparNet, a convolutional neural network composed of five parallel convolutional filters and the SENet, is proposed to learn EEG space-frequency domain characteristics and distinguish between depressive and normal control. The model is trained and tested by the cross-validation method of subject division. The results show that SparNet achieves a sensitivity of 95.07%, a specificity of 93.66%, and an accuracy of 94.37% in classification. Therefore, our results can conclude that the proposed SparNet model is effective in detecting depression using EEG signals. It also indicates that the combination of spatial information and frequency domain information is an effective way to identify patients with depression.

Tendencias y perspectivas en la detección de depresión mediante el análisis de señales Electroencefalográficas (EEG): una Revisión Sistemática de la Literatura.

“…Después del preprocesamiento, se emplearon las arquitecturas de Aprendizaje Profundo para subrayar el rendimiento de discriminación comparando precisiones de clasificación de lo cual se obtuvo una precisión del 89,33% y 92,66% con arquitectura MobileNet para el hemisferio izquierdo y derecho, respectivamente [Uyulan et al, 2021]. Las Redes Neuronales han destacado por ser muy utilizadas para detectar depresión mediante señales EEG, por ejemplo, un estudio con una muestra con 30 sujetos sanos y 34 con depresión, utilizó Redes Neuronales Convolucionales de una y dos dimensiones (1DCNN-2-DCNN), LSTM, una combinación de CNN (del inglés, Convolutional Neural Networks) con el modelo LSTM (1DCNN LSTM y 2DCNN-LSTM) obteniendo una precisión del 99.24% [Saeedi et al, 2020], otro estudio obtuvo el 94.3% con una muestra de 154 sujetos donde el modelo se entrenó mediante una validación cruzada de división de sujetos [Xin et al, 2022]. También se encontró otro estudio basado en Redes Neuronales Convolucionales compuestas por 4 bloques: convolución, convolución en profundidad, convolución separable y clasificación.…”

Enhancing generalized anxiety disorder diagnosis precision: MSTCNN model utilizing high-frequency EEG signals

“…At present, the attention mechanism has gradually become a boom in deep learning, and an increasing number of researchers are applying it to EEG signal processing. Deng et al ( 57 ) improved the accuracy of major depressive disorder classification from 91.24% to 94.37% by adding SE attention mechanism to one-dimensional convolution. Chen et al used CBAM attention for ResNet34 in emotion recognition task, and the accuracy increased by 5.54% compared with ResNet34 ( 58 ).…”

Depression assessment using integrated multi-featured EEG bands deep neural network models: Leveraging ensemble learning techniques

“…The EEG signals are mainly 4 frequency bands using 19 electrodes from 46 normal and 46 depressed subjects. In the literature [48] , a CNN-based SparNet depression discrimination model was proposed to learn EEG space-frequency domain characteristics. The study involved the collection of EEG signals from 48 subjects, with 24 normal and 24 depressed subjects.…”