Efficient TMS-Based Motor Cortex Mapping Using Gaussian Process Active Learning

Abstract

Transcranial Magnetic Stimulation (TMS) can be used to map cortical motor topography by spatially sampling the sensorimotor cortex while recording Motor Evoked Potentials (MEP) with surface electromyography (EMG). Traditional sampling strategies are time-consuming and inefficient, as they ignore the fact that responsive sites are typically sparse and highly spatially correlated. An alternative approach, commonly employed when TMS mapping is used for presurgical planning, is to leverage the expertise of the coil operator to use MEPs elicited by previous stimuli as feedback to decide which loci to stimulate next. In this paper, we propose to automatically infer optimal future stimulus loci using active learning Gaussian Process-based sampling in place of user expertise. We first compare the user-guided (USRG) method to the traditional grid selection method and randomized sampling to verify that the USRG approach has superior performance. We then compare several novel active Gaussian Process (GP) strategies with the USRG approach. Experimental results using real data show that, as expected, the USRG method is superior to the grid and random approach in both time efficiency and MEP map accuracy. We also found that an active warped GP entropy and a GP random-based strategy performed equally as well as, or even better than, the USRG method. These methods were completely automatic, and succeeded in efficiently sampling the regions in which the MEP response variations are largely confined. This work provides the foundation for highly efficient, fully automatized TMS mapping, especially when considered in the context of advances in robotic coil operation.

M2M-InvNet: Human Motor Cortex Mapping from Multi-Muscle Response Using TMS and Generative 3D Convolutional Network

“…In future work, we plan to include cortical motor topography mapping using active learning [23] and to study the generation of the volume conductor model by deep learning, to determine if we can combine these with the current expert user-guided mapping and the segmentation-finite element simulation pipeline, respectively, or perhaps even replace either or both entirely.…”

“…The details of these preprocessing techniques are outlined in Section II-C. For each map, TMS (100-300 stimulations, 4 s ISI) was delivered over a 6×6 cm regular grid (1 cm spacing, 36 cm 2 area) centered on the hotspot. For each intensity, one stimulus was delivered to each of the 49 equidistant points on the grid, and the remaining stimulations were delivered using real-time feedback from the MEPs to maximize information about the responsive areas [5], [22], [23]. Care was taken to ensure that the mapping included the full extent of the excitable area for all recorded muscles.…”

“…The remaining stimuli (51-251 per intensity for subject 1 and 250-251 per intensity for subjects 2 and 3) were delivered within the 6×6 cm area defined by the grid at loci selected by the expert TMS operator using real-time feedback from the MEPs to maximize information about the responsive areas. We have previously shown that this technique produces similar information to traditional gridded mapping approaches [24]. Care was taken to ensure that the mapping included the full extent of the excitable area at the given stimulation intensities for all recorded muscles.…”

Gaussian-process-based Bayesian optimization for neurostimulation interventions in rats

“…As demonstrated in the primary publication 1 and elsewhere, 2 , 3 , 4 , 5 , 6 Gaussian-process (GP)-based Bayesian optimization (BO) algorithms are a powerful framework to automatically optimize the efficacy of neurostimulation. It has been shown to outperform other strategies to simultaneously find the optimal values of multiple stimulation parameters (i.e., the optimal combination of parameter values) to maximize a chosen feature of the evoked response (e.g., the movement amplitude or the electromyographic [EMG] burst amplitude).…”

M2M-InvNet: Human Motor Cortex Mapping From Multi-Muscle Response Using TMS and Generative 3D Convolutional Network

“…The remaining stimuli (51-251 per intensity for subject 1 and 250-251 per intensity for subjects 2 and 3) were delivered within the 6x6 cm area defined by the grid at loci selected by the expert TMS operator using real-time feedback from the MEPs to maximize information about the responsive areas. We have previously shown that this technique produces similar information to traditional gridded mapping approaches [24]. Care was taken to ensure that the mapping included the full extent of the excitable area at the given stimulation intensities for all recorded muscles.…”