Schedule of Events | Symposia

Resting-State EEG Dynamics and Neurophysiological Mechanisms of Transcranial Direct Current Stimulation Responses in Neuropsychiatric Disorders

Poster Session A - Saturday, March 29, 2025, 3:00 – 5:00 pm EDT, Back Bay Ballroom/Republic Ballroom

Katerina Nasto¹ (knasto@mgh.harvard.edu), Samadrita Chowdhury², Asif Jamil², Shane Walsh¹, Joan Camprodon²; ¹Massachusetts General Hospital, ²Massachusetts General Hospital, Harvard Medical School

Background: Transcranial direct current stimulation (tDCS) is a promising neuromodulation approach for addressing cognitive deficits in neuropsychiatric disorders. However, outcomes show significant heterogeneity, partially due to patient-specific differences in pathophysiology and/or normal neurobiology, highlighting the need for precision approaches. Variations in baseline brain states, captured through resting-state EEG, may explain differential responses. This study investigates neurophysiological signatures of tDCS efficacy, emphasizing their role in efficient neural processing and cognitive control. Methods: We applied unsupervised machine learning (ML) to stratify tDCS outcomes based on resting-state EEG in individuals with ADHD, Substance Use Disorders, and healthy controls (n=102). Participants underwent tDCS targeting executive function circuits, with pre-/post-stimulation behavioral task assessments (Flanker, N-Back, Stop Signal) and EEG. Baseline EEG profiles were created by extracting spectral features, including relative/absolute power, entropy, and the spectral exponent (1/f slope of power spectral density). Similarity Network Fusion (SNF) and spectral clustering identified EEG-based clusters. Correlations between clusters and post-tDCS outcomes were assessed, with EEG feature contributions evaluated via normalized mutual information and ANOVA. Results: The two EEG-based clusters showed significant differences in behavioral response (p=0.0028, p=4.8e-5) and electrophysiology (p<2.2e-16). The responder cluster exhibited higher frontal alpha power and a steeper 1/f slope, indicative of optimized excitatory-inhibitory balance and cognitive control. ML-derived clusters outperformed diagnostic categories in predicting tDCS response (p=0.018), highlighting the value of data-driven stratification. Conclusion: Our approach identified distinct neurophysiological signatures of tDCS outcomes, advancing our understanding of the intervention’s mechanisms and highlighting the potential of ML-based and neurobiology-informed methods for personalized neuropsychiatric interventions.

Topic Area: EXECUTIVE PROCESSES: Other