Schedule of Events | Symposia

Precision Neuromodulation using closed-Loop TMS/fMRI and Neural Network-Based Brain State Decoding

Poster Session B - Sunday, March 30, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Ahsan Khan¹ (ahsan.khan@pennmedicine.upenn.edu), Hongming Li¹, Camille Blaine¹, Julie Grier¹, Ethan Hammet¹, Almaris Figueroa¹, Sarai Garcia¹, Romain Duprat¹, Justin Reber¹, Joseph Deluisi¹, Yong Fan¹, Desmond Oathes¹; ¹Perelman School of Medicine, University of Pennsylvania

Transcranial magnetic stimulation (TMS) has transformed non-invasive brain therapies but faces challenges due to variability in outcomes, likely stemming from individual differences in brain structure and function. This study aims to address this challenge by integrating individualized functional connectivity networks (FCNs) derived from functional magnetic resonance imaging (fMRI) with a neural network-based decoder for adaptive, real-time stimulation delivery. We investigated how TMS targeting a brain region associated with working memory affects task performance over several days of neuromodulation in healthy participants. After collecting behavioral measures and screening questionnaires in the first two visits, participants underwent a concurrent TMS/fMRI session on the third visit. Real-time decoder readouts during this session identified stimulation frequencies classified as optimal or suboptimal based on their effects on brain activity and behavior. Participants then underwent a crossover design, receiving three optimal and three suboptimal stimulation sessions (six total) in randomized order. During each session, participants completed both a working memory task and a control task. Post-training TMS/fMRI sessions (visits 7 and 11) assessed the effects of stimulation and decoder performance. Results showed improved working memory performance during optimal stimulation compared to suboptimal stimulation, with no changes observed in the control task. Further analysis showed a consistent negative correlation between the brain readouts (larger value indicates more engaged in working memory task) and reaction time and error rates during the working memory task. This highlights the potential of real-time brain decoding to enhance TMS efficacy and pave the way for precise, personalized neuromodulation.

Topic Area: EXECUTIVE PROCESSES: Working memory