Development of an AI-driven robotic TMS system for versatile, high-resolution cortical mapping

PROJECT SUMMARY / ABSTRACT Transcranial magnetic stimulation (TMS) is a powerful, non-invasive technique for probing and modulating human brain circuits. However, most TMS systems rely on manual coil positioning, which introduces spatial inaccuracy (often exceeding 5-11 mm across sessions) and operator dependence, compromising the reliability of cortical mapping. These limitations are particularly problematic because millimeter mishits can convert treatment responders to non-responders in individualized neuromodulation protocols. This project aims to address these challenges by developing an AI-guided robotic TMS system that can automatically position and adjust the coil in real time based on feedback from neural responses. The proposed system integrates a six-degree-of-freedom (6-DoF) robotic arm for high-resolution coil positioning and a closed-loop Gaussian-process Bayesian optimization algorithm that iteratively optimizes coil positioning using motor-evoked potential (MEP) feedback. This engineering-focused approach is designed to achieve ±2 mm spatial and ±2° orientation precision, improving spatial accuracy by an order of magnitude and reducing stimulation requirements by 4-5 times compared to conventional manual methods. The project has three specific aims: (1) to develop the robotic TMS coil holder with integrated safety systems and brain phantoms for validation; (2) to implement the Bayesian optimization algorithm that reduces stimuli from ~50-100 to ≤17 pulses for motor hotspot localization; and (3) to validate cortical mapping performance using sophisticated brain phantoms that simulate MEP responses. Phantom-based validation enables comprehensive system characterization under controlled conditions while establishing the foundation for future human studies applications. This NIH R15 AREA project will provide immersive research training for three undergraduate students per year at the New Jersey Institute of Technology, a primarily undergraduate institution. Students will rotate through all project phases, including hardware design, algorithm development, phantom validation studies, and dissemination. The expected outcomes include a validated open-source platform, six conference presentations, three publications, and comprehensive documentation enabling widespread adoption across institutions. By combining neuroengineering, robotics, and adaptive control with rigorous phantom validation, this work will advance next-generation autonomous TMS technologies and contribute to training the next generation of neurotechnology workforce.