(15) Comparing Thalamic Atlases and Segmentation Techniques used in Deep Brain Stimulation Modeling for Essential Tremor

Introduction: Deep brain stimulation (DBS) of the ventral intermediate (VIM) nucleus of the thalamus is a proven treatment for essential tremor (ET). Neurosurgeons rely on thalamic atlases to locate relevant nuclei for stimulation; however, many atlases are available, with different divisions of motor and sensory subnuclei[1]. Segmenting anatomical targets using diffusion tensor imaging (DTI) offers a patient-specific alternative to traditional atlas-based approaches[2]. This work compares the effectiveness of multiple atlases and segmentation techniques in defining motor and sensory subnuclei, evaluated based on volume of tissue activation (VTA) correspondence with clinical outcomes.

Methods: 22 ET patients who underwent unilateral VIM DBS were analyzed retrospectively[2]. Six prominent, dissimilar thalamic atlases were obtained from LEAD-DBS (Fig. 1)[3]. After registering each atlas to each patient thalamus, two segmentation approaches were used: atlas-based segmentation (ABS) and DTI-based segmentation (DTIBS). In ABS, patient thalamus voxels were classified by the atlas subnucleus they resided in; in DTIBS, a k-means clustering algorithm classified voxels using spatial and tissue properties. Patient thalami were aggregated, and motor/sensory-VTA overlap between techniques was analyzed using a paired, two-sided Wilcoxon signed-rank test (Fig. 2).

Results: For VTAs associated with tremor reduction, ABS showed significantly greater motor than sensory activation in 6/6 atlases, while DTIBS showed greater motor than sensory activation in 3/6 atlases. Moreover, sensory activation was significantly greater for paresthesia VTAs than for therapeutic VTAs in 5/6 atlases using ABS. Contrastingly, DTIBS had greater sensory overlap for paresthesia VTAs than tremor reduction VTAs in 3/6 atlases. Geometrically, ABS produced a significantly larger motor volume (p=3.62E-9) and smaller sensory volume (p=9.31E-8) than DTIBS when grouping all patients and atlases.

Conclusion: High motor and low sensory activation is expected for tremor reduction stimulation, and ABS showed this trend across all atlases. In general, ABS produced a larger motor subnuclei volume and smaller sensory subnuclei volume than DTIBS. The larger motor volume of ABS increased VTA-motor overlap, contributing to the greater motor activation for therapeutic outcomes seen in ABS. Greater sensory stimulation is expected for paresthesia VTAs compared to tremor reduction VTAs, and ABS yielded this trend in almost all atlases.
ABS was a more generalizable segmentation method than DTIBS across atlases, as the resulting subnuclei from ABS aligned more closely with the observed clinical outcomes. The Jakab atlas[4] was the only atlas to show the expected motor/sensory stimulation trend associated with each clinical outcome when using either ABS or DTIBS, making it the most clinically accurate atlas investigated.