Targeting multi-locus TMS with Bayesian optimization

Abstract

This approach ensures that stimulation is directed toward relevant neural pathways, improving the precision of dual-site TMS.Synchronization of paired-pulse stimulation across both sites is tightly controlled, ensuring accurate interstimulus intervals (ISIs).In a pilot experiment, motor evoked potentials were recorded after stimulation of the abductor pollicis brevis (APB) muscles in both hemispheres, with the robotic systems maintaining both coils position.Integrating machine learning and tractography enabled more accurate and automated identification of stimulation targets, further refining the targeting process.Compared to manual methods, our system demonstrated superior accuracy (0.3 mm in distance and 0.2 in angle deviations) and reduced variability.This innovative approach minimizes reliance on operator expertise and enhances stimulation reproducibility.By combining machine learning, tractography, and robotic precision, this dual-site TMS system enables advanced brain stimulation techniques, including motor mapping, hotspot identification, and network-based stimulation protocols, providing a more effective and reliable platform for both research and clinical applications.