Item

Abstract

Transcranial magnetic stimulation (TMS) is a non-invasive brain stimulation tool for modulating and mapping cortical function. TMS-based functional mapping methods typically exploit causal relationships between brain structure and function. In earlier work, we demonstrated that cortical sites of the motor area can be precisely localized by relating FEM simulation derived electric field estimates to TMS induced motor evoked potentials (MEPs) in an input-output curve fitting approach using nonlinear regression. A random sequence of on average 180 TMS coil configurations (= coil position and orientation) have proven to provide sufficient variability in their electric field pattern and evoked motor responses to discriminate neuronal populations at the primary motor cortex at single digit level based on their goodness of fit.

Here, we propose an optimized robotic TMS mapping method to significantly reduce the number of stimulation samples, improve the mapping result and reduce the time required for the mapping experiment. Starting from a set of random TMS coil configurations, new configurations are determined during the experiment based on two Pareto-weighted criteria: a) the correlation of the induced electric fields is minimized to increase the distinguishability between cortical locations, and b) the information gain of the candidate position is maximized by determining the optimal sample location on the I/O curve using Fisher information matrix optimization informed by real-time MEP feedback.

The presented method was implemented in a robotic TMS system to enable accurate and reliable coil positioning, exactly timed triggering of the TMS pulse, and the online acquisition and analysis of the MEPs.