EEG-ViLSTM: A Deep Learning Approach for Depression Detection Using EEG Signals

Depression is a global mental health disorder that significantly affects the quality of life for those experiencing it. Early detection and accurate diagnosis are therefore crucial for timely intervention and treatment to improve the quality of life. State-of-the-art deep learning models have been helpful for the early detection of symptoms, especially through the analysis of electroencephalogram (EEG) signals. However, due to the variability in these signals across individuals, these deep learning-based models fail to generalize across diverse populations potentially leading to inaccurate or inconsistent results. To address this challenge, this work presents EEG-ViLSTM, a novel deep learning-based approach that employs Vision-LSTM and operates on EEG signals. EEG-ViLSTM sets a new benchmark for depression classification with an accuracy of 93.52% with 0.94 precision and a recall and F1-score of 0.93 on the multi-modal open dataset for mental disorder analysis (MODMA) dataset. Experimental results demonstrate the effectiveness of EEG-ViLSTM and claim the superior detection capability over various state-of-the-art techniques and indicate its utility as a clinical tool for depression evaluation.