Probabilistic comparison of gray and white matter coverage between depth and surface intracranial electrodes in epilepsy

Abstract

In this study, we quantified the coverage of gray and white matter during intracranial electroencephalography in a cohort of epilepsy patients with surface and depth electrodes. We included 65 patients with strip electrodes (n = 12), strip and grid electrodes (n = 24), strip, grid, and depth electrodes (n = 7), or depth electrodes only (n = 22). Patient-specific imaging was used to generate probabilistic gray and white matter maps and atlas segmentations. Gray and white matter coverage was quantified using spherical volumes centered on electrode centroids, with radii ranging from 1 to 15 mm, along with detailed finite element models of local electric fields. Gray matter coverage was highly dependent on the chosen radius of influence (RoI). Using a 2.5 mm RoI, depth electrodes covered more gray matter than surface electrodes; however, surface electrodes covered more gray matter at RoI larger than 4 mm. White matter coverage and amygdala and hippocampal coverage was greatest for depth electrodes at all RoIs. This study provides the first probabilistic analysis to quantify coverage for different intracranial recording configurations. Depth electrodes offer increased coverage of gray matter over other recording strategies if the desired signals are local, while subdural grids and strips sample more gray matter if the desired signals are diffuse.

Figure 1

Overview of Methods. (A) The processing pipeline used patient specific imaging. All patients were processed through LeGUI, FreeSurfer, and SIMNIBS, while only surface electrode patients received extra processing to project the electrodes to the smoothed gray matter surface as well as left and right hemisphere segmentation for interhemispheric contacts. Overlap of gray and white matter segmentations with RoIs were used to calculate volumes of gray and white matter coverage. A subset of patients were used to run FEM-based RoIs. Reported sample sizes indicate the number of patients included in each implant category. (B) Electrode localization was performed in the freely available LeGUI software to mark strip, grid, and depth electrodes across a cohort of 65 patients. LeGUI automatically processes probabilistic gray and white matter segmentations and numerous atlas registrations including the AAL and NMM atlases. (C) Substantial brain shift is frequently observed in subdural electrode cases during surgical implantation. After image co-registration, the subdural electrodes localized from the CT may appear inside of the brain (yellow) rather than on top of the cortical surface. (D) All subdural electrodes were projected (yellow—original location, black—projected location) onto the smooth gray matter surface according to Hermes et al., 2010. (E) In patients with interhemispheric subdural electrodes, the left and right hemisphere segmentations were created in FreeSurfer so that volumes would be restricted to the hemisphere from which the electrode recorded neural activity. (F) Spheres at fixed radii from 1–15 mm at 0.5 mm steps were placed at the contact centroids. 10 mm RoIs are shown in blue. Overlap of spherical volumes with gray and white matter segmentations quantify the amount of gray and white matter coverage for different modalities.

Figure 2

Gray and white matter coverage per contact across RoIs. (A) There was a dynamic relationship between gray matter coverage and the size of the RoI used across intracranial modalities. (B) Across notable RoIs at 2.5 mm, 5 mm, 10 mm, and 15 mm, depth electrodes covered more gray matter at small RoIs than strip electrode or strip and grid electrode combinations (S & D: p = 3.2 × 10^–5 ; S + G & D: p = 5.34 × 10^–5); however, the relationship reversed for strip + grid and depth electrodes at a 5 mm RoI (S + G & D: p = 0.00326) and reversed for strip and depth electrodes at a 10 mm RoI (S & D: p = 1.86 × 10^–5). At 15 mm, strip electrodes exceeded other intracranial electrode approaches in gray matter coverage on a per contact basis (S & S + G: p = 5.17 × 10^–7; S & S + G + D: p = 3.60 × 10^–5; S & D: p = 0.00477). (C) White matter coverage varied across recording modalities, but was dominated by depth electrodes. Though not significant, cases that used strip, grid, and depth electrodes in combination appeared to increase the white matter sampling over cases that used subdural electrodes alone at RoIs less than 12 mm. (D) Depth electrodes sample more white matter than configurations that used subdural electrodes. (2.5 mm—S & D: p = 1.61 × 10^–5 ; S + G & D: p = 3.84 × 10^–9; 5 mm—S & D: p = 1.94 × 10^–6 ; S + G & D: p = 4.45 × 10^–9; 10 mm—S & D: p = 9.58 × 10^–5 ; S + G & D: p = 4.02 × 10^–9; 15 mm—S & D: p = 0.00768; S + G & D: p = 3.78 × 10^–9; S + G + H & D: p = 0.00446) E. Relative to depth electrodes, subdural electrodes covered less gray matter prior to a 4 mm RoI. At larger radii, subdural electrodes covered more gray matter until a RoI of 12 mm when only the depth electrodes surpassed configurations strip and grid configurations and strip, grid, and depth configurations. (F) Across all RoIs, depth electrodes covered more white matter. White matter coverage for very small RoIs was close to zero, while at very large RoIs, depth electrodes covered approximately twice the volume of white matter as subdural electrodes.

Figure 3

Region-specific coverage of gray matter was calculated across all patients. (n) on the x-axis label indicates the number of patients included. Images of brain regions were created in SCIRun5 version beta.Y (from the Scientific Computing and Imaging Institute, SCI) (A) All patients had some coverage of the frontal and temporal lobes regardless of the modality used. (B) Notably, hippocampal and amygdala coverage of depth electrodes greatly exceeded coverage using subdural electrode configurations. (C) Patients with depth contacts only had significantly more insula coverage than patients with subdural strip contacts. Nearly all depth electrode only patients had coverage in the insula and cingulate cortex.

Figure 4

Use of patient-specific brain modeling to create RoIs. (A) Using a head mesh generated from a FreeSurfer/SIMNIBS pipeline. The finite element method was used to solve for voltage solutions for all electrodes for 3 patients (1 strip patient, 1 strip and grid patient, 1 depth patient). FEM-based RoI volumes were determined by equipotential lines according to the average voltage at 2.5 mm, 5 mm, 10 mm, and 15 mm away. (B) Using the Dice coefficient to measure overlap, FEM-based volumes were not statistically significantly different in shape to spherical volumes. However, depth electrodes had significantly higher dice coefficients than surface electrodes at all RoI sizes (2.5 mm—S & S + G: p = 5.44 × 10^–5 ; S & D: p = 3.09 × 10^–12 ; S + G & D: p = 0; 5 mm—S & S + G: p = 1.54 × 10^–8; S & D: p = 0; S + G & D: p = 7.21 × 10^–10; 10 mm—S & S + G: p = 5.82 × 10^–3; S & D: p = 5.35 × 10^–15 ; S + G & D: p = 1.05 × 10^–7; 15 mm—S & D: p = 5.82 × 10^–6; S + G & D: p = 0.0013). (C) Slight differences in individual FEM-based volumes did not translate to significant differences in the calculated coverage of gray matter. (D) Overall, across patients and all RoIs, there was no consistent pattern in how total gray matter coverage changed with the use of a FEM-based RoI, thus, justifying the use of spherical RoIs in our model.