Functional brain network modulation following conventional and optimized HD-tDCS in major depressive disorder: A machine learning prediction of treatment response

Mikaeil Tabarraei; Sepideh Jabbari; Marzieh Abdi; Mohsen Dadashi; Mohammad Ali Salehinejad

doi:10.1016/j.jad.2026.121595

Functional brain network modulation following conventional and optimized HD-tDCS in major depressive disorder: A machine learning prediction of treatment response

J Affect Disord. 2026 Aug 1:406:121595. doi: 10.1016/j.jad.2026.121595. Epub 2026 Mar 9.

Authors

Mikaeil Tabarraei¹, Sepideh Jabbari², Marzieh Abdi³, Mohsen Dadashi⁴, Mohammad Ali Salehinejad⁵

Affiliations

¹ Electrical Engineering Department, Engineering Faculty, University of Zanjan, Zanjan, Iran.
² Electrical Engineering Department, Engineering Faculty, University of Zanjan, Zanjan, Iran. Electronic address: sjabbari@znu.ac.ir.
³ Department of Clinical Psychology, Faculty of Medicine, Zanjan University of Medical Sciences, Zanjan, Iran.
⁴ Department of Clinical Psychology, Faculty of Medicine, Zanjan University of Medical Sciences, Zanjan, Iran. Electronic address: mohsendadashi@zums.ac.ir.
⁵ Department of Child and Adolescent Psychiatry, Psychosomatics and Psychotherapy, Medical Faculty, RWTH Aachen University, Aachen, Germany.

PMID: 41812893
DOI: 10.1016/j.jad.2026.121595

Abstract

Background: Major depressive disorder (MDD) is a prevalent and disabling condition that remains inadequately treated in many patients. Transcranial direct current stimulation (tDCS) is a non-invasive intervention with evidence for mood improvement, and high-definition tDCS (HD-tDCS) offers more focal stimulation. This study compared functional brain connectivity patterns between MDD patients and healthy controls, and evaluated how 4 mA optimized HD-tDCS, 2 mA conventional tDCS, and sham interventions altered connectivity in the MDD group.

Methods: We analyzed resting state EEG data of 60 patients with MDD who completed 30 sessions of conventional tDCS (n = 20), optimized HD-tDCS (n = 20), or sham stimulation. A separate EEG dataset from 61 demographically matched healthy participants served as a reference dataset (49 analyzed after quality control). Functional connectivity metrics derived from EEG signals were compared across and within stimulation protocols and between responders and nonresponders.

Results: Compared to healthy controls, patients showed reduced baseline connectivity. Following active tDCS, responders showed significant connectivity increases that correlated with reduced depression severity. Specifically, responders demonstrated enhanced network strength, global efficiency, and local efficiency post-intervention. A machine learning model using PCA-based dimensionality reduction and participant-level validation predicted treatment response with 81% accuracy. Connectivity gains were significantly greater following active tDCS, with optimized HD-tDCS yielding higher post-treatment connectivity than both conventional and sham protocols.

Conclusion: These findings suggest that optimized HD-tDCS may be a more effective therapeutic option for MDD by more effectively modulating brain network communications.

Keywords: Depression; Functional connectivity; Graph theory; HD-tDCS; Machine learning; tDCS.

MeSH terms

Adult
Brain* / physiopathology
Electroencephalography
Female
Humans
Machine Learning*
Major Depressive Disorder* / physiopathology
Major Depressive Disorder* / therapy
Male
Middle Aged
Transcranial Direct Current Stimulation* / methods
Treatment Outcome