Schedule of Events | Symposia

Real-Time Modulation of Reinforcement Learning Using Closed-Loop TMS-EEG

Poster Session B - Sunday, March 30, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom
Also presenting in Data Blitz Session 4 - Saturday, March 29, 2025, 10:30 am – 12:00 pm EDT, Constitution B.

Yifan Gao¹ (yifangao97@gmail.com), Malte Güth², Drew Headley¹, Daniel Robles¹, Emily Zhang¹, Travis Baker¹; ¹Center for Molecular and Behavioral Neuroscience, Rutgers University – Newark, ²Department of Biomedical Engineering, University of Minneapolis, Minnesota

Background: While the link between frontal-midline theta (FMT) power, reward prediction errors, and behavioral adaption is well established, little is known about the role of FMT phase dynamics in reinforcement learning. Here, we developed a novel closed-loop system capable of tracking FMT in real-time and tested whether precisely triggering TMS synchronized with the peak or trough of FMT following feedback would impact the electrophysiological and behavioral correlates of reinforcement learning. Methods: Thirty-four participants were randomly assigned to either peak or trough closed-loop stimulation, and completed two sessions (active and sham). For each session, participants completed two decision-making tasks and received peak or trough stimulation following positive and negative feedback. The T-maze task assessed phase effects on neural responses to feedback (ERPs), while the Probabilistic Selection Task (PST) evaluated the effects on learning. Results: Relative to sham (M=2.48μV), trough stimulation diminished the reward positivity (M=0.79μV, t[16]=-2.42, p=0.028), an ERP component associated with reward processing. While the ability to generalize learning to novel pairing was spared across TMS conditions, the accuracy of learned stimulus-response mappings was impaired by peak stimulation (M=0.76) relative to sham (M=0.87, t[16]=-2.43, p=0.028). Further, trough stimulation slowed reaction time (M=979ms) compared to sham (M=877ms, t[16]=2.31, p=0.035). Discussion: These findings highlight the impact of phase-specific FMT stimulation on neural reward responses and stimulus-response associations, providing new insights into the function of FMT phase dynamics in reinforcement learning, as well as a therapeutic TMS target for cognitive impairments in conditions like substance use disorders, schizophrenia, ADHD, and traumatic brain injury.

Topic Area: THINKING: Decision making