Segmentation of the Globus Pallidus Internus Using Probabilistic Diffusion Tractography for Deep Brain Stimulation Targeting in Parkinson Disease

Abstract

Background and purpose: Although globus pallidus internus deep brain stimulation is a widely accepted treatment for Parkinson disease, there is persistent variability in outcomes that is not yet fully understood. In this pilot study, we aimed to investigate the potential role of globus pallidus internus segmentation using probabilistic tractography as a supplement to traditional targeting methods.

Materials and methods: Eleven patients undergoing globus pallidus internus deep brain stimulation were included in this retrospective analysis. Using multidirection diffusion-weighted MR imaging, we performed probabilistic tractography at all individual globus pallidus internus voxels. Each globus pallidus internus voxel was then assigned to the 1 ROI with the greatest number of propagated paths. On the basis of deep brain stimulation programming settings, the volume of tissue activated was generated for each patient using a finite element method solution. For each patient, the volume of tissue activated within each of the 10 segmented globus pallidus internus regions was calculated and examined for association with a change in the Unified Parkinson Disease Rating Scale, Part III score before and after treatment.

Results: Increasing volume of tissue activated was most strongly correlated with a change in the Unified Parkinson Disease Rating Scale, Part III score for the primary motor region (Spearman r = 0.74, P = .010), followed by the supplementary motor area/premotor cortex (Spearman r = 0.47, P = .15).

Conclusions: In this pilot study, we assessed a novel method of segmentation of the globus pallidus internus based on probabilistic tractography as a supplement to traditional targeting methods. Our results suggest that our method may be an independent predictor of deep brain stimulation outcome, and evaluation of a larger cohort or prospective study is warranted to validate these findings.

Fig 1.

Statistical heat maps illustrating the group average location of each of the 10 GPi segments. Maps are thresholded at >10%. PPN indicates pedunculopontine nucleus.

Fig 2.

Scatterplots showing the change in the UPDRS score relative to volumes of tissue activated in the primary motor segment (upper left), supplementary motor area/premotor cortex segment (upper right), globus pallidus externus segment (lower left), and substantia nigra segment (bottom right).

Fig 3.

Left (A) and right (D) group average volumes for segments with maximal connectivity to the primary motor cortex (yellow) and supplementary motor area/premotor cortex (light blue). B, Left DBS leads and volumes of tissue activated (red) for the 3 patients with improvement in the UPDRS III of >10 (mean improvement in the score = 26.7) after DBS. A more ventral location and greater overlap with the primary motor cortex segment can be appreciated. C, Left DBS leads and volumes of tissue activated (red) in the 3 patients with the least improvement in the UPDRS III score (mean improvement in the score = 0.7) show a more dorsal location of the volume of tissue activated with less overlap with the primary motor cortex segment of the GPi. E, Right DBS leads and volumes of tissue activated. All patients had an improvement in the UPDRS III score of >10 (mean improvement in the score = 19).