Depression detection and subgrouping by using the active and passive EEG paradigms

Depression, a paramount global health challenge, necessitates an advanced diagnostic approach. This study employs EEG and AI on a psycho-physiological Healthy Brain Network (HBN) database of 100 patients to recognize and subgroup major depressive disorders during diverse cognitive tasks. After denoising EEG signals and extracting 16383 features, the feature space’s dimensionality is reduced using a minimal-redundancy-maximal-relevance technique. Various AI classifiers (SVM, DT, NB, LR) discern depressed patients from normal controls, and hierarchical cluster analysis (HCA) with t-SNE and kernel PCA showcases superior clustering indices in both active and passive tasks. The results highlight the active sequence learning paradigm’s efficacy, achieving 100% accuracy with a decision tree classifier, underscoring its relevance in discerning depression and revealing its influence on cognitive abilities related to memory and learning. This comprehensive approach not only accentuates the active sequence learning paradigm’s exceptional efficacy with two subgroups but also underscores its pivotal role in unraveling the intricate associations between depression, memory, and learning capacities.