Skip to content

Ensemble of XGBoost for Depression Level Classification using EEG Functional Connectivity in Mindfulness Based Cognitive Therapy (MBCT)

E. Agustin, S. W. Purnami, W. Islamiyah, M. M. Maramis, S. Karimah

International Conference Advancement Data Science, E-learning and Information Systems June 1, 2026 DOI: 10.1109/icadeis71120.2026.11644519 (opens in new tab)

Relevant Not relevant

Study at a glance

AI-extracted from the abstract
Characteristics Observational study with machine learning classification Peer reviewed
Sample size 32
Population 12 normal subjects and 20 depression subjects
Intervention Mindfulness Based Cognitive Therapy (MBCT)
Topics Depression Meditation
Key findings The EoXGBoost model classified depression from EEG functional connectivity features with cross-validated accuracy of 98.80% and AUC of 99.97%. Segment-level predictions indicated temporal variability in depression states within individuals. After Mindfulness Based Cognitive Therapy, the model predicted approximately 27% improvement without worsening, but the authors report that these changes were not statistically significant.

Abstract

Depression is a mental health disorder that requires accurate detection and effective treatment. Electroencephalography (EEG) based analysis has emerged as a promising approach for objective depression evaluation. This study leverages selected EEG features using hybrid feature selection approach to train an EoXGBoost model and evaluates its generalization on unlabeled posttest depression data. In addition, this research focuses on capturing temporal variability in depression-related EEG patterns across segments. A total of 32 subjects were included in this study, including 12 normal subjects and 20 depression subjects. Functional connectivity features were extracted using Pearson Correlation, Mutual Information, and Phase Lag Index across multiple channel pairs and frequency subbands, resulting in high-dimensional data. To address this issue, a hybrid feature selection approach was applied to reduce the feature dimensionality while preserving strong classification performance with low computational cost. Model performance was evaluated using nested cross validation with evaluation metrics including accuracy (98.80%), sensitivity (98.76%), precision (98.89%), specificity (99.34%), F1-score (98.82%), and AUC (99.97%). The trained model was further applied to classify unlabeled posttest EEG data. Segment level classification revealed temporal variability in depression states within individuals, and a majority voting strategy was used to determine subject level classification. The framework is then utilized to analyze predicted changes in depression levels following Mindfulness Based Cognitive Therapy (MBCT) treatment. The results demonstrate that the proposed approach not only achieves accurate classification but also provides insight into temporal dynamics and potential changes after MBCT treatment with approximately 27% predicted improvement without worsening. However, statistical analysis revealed that the observed changes following MBCT were not statistically significant.

Comparable studies

Other observational and cohort studies on depression, most cited first.

Study Year Design Participants
Metacognitive awareness and prevention of relapse in depression: Empirical evidence. Residually depressed patients and nondepressed controls; recovered depressed patients 2002 Observational cohort with intervention components
Psilocybin for treatment-resistant depression: fMRI-measured brain mechanisms. Patients with treatment-resistant depression 2017 Observational cohort n = 19
Antidepressant effects of a single dose of ayahuasca in patients with recurrent depression: a preliminary report Patients with a depressive disorder 2015 Observational study
Concomitant BDNF and sleep slow wave changes indicate ketamine-induced plasticity in major depressive disorder Patients with treatment-resistant major depressive disorder 2012 Observational cohort n = 30
Altered peripheral immune profiles in treatment-resistant depression: response to ketamine and prediction of treatment outcome Healthy controls and actively depressed patients with treatment-resistant depression... 2017 Observational cohort n = 59

Citations in the library

Cites 2

Related reading

Explore topics

Depression Meditation