Local functional connectivity as a pre-surgical tool for seizure focus identification in non-lesion, focal epilepsy

Abstract

Successful resection of cortical tissue engendering seizure activity is efficacious for the treatment of refractory, focal epilepsy. The pre-operative localization of the seizure focus is therefore critical to yielding positive, post-operative outcomes. In a small proportion of focal epilepsy patients presenting with normal MRI, identification of the seizure focus is significantly more challenging. We examined the capacity of resting state functional MRI (rsfMRI) to identify the seizure focus in a group of four non-lesion, focal (NLF) epilepsy individuals. We predicted that computing patterns of local functional connectivity in and around the epileptogenic zone combined with a specific reference to the corresponding region within the contralateral hemisphere would reliably predict the location of the seizure focus. We first averaged voxel-wise regional homogeneity (ReHo) across regions of interest (ROIs) from a standardized, probabilistic atlas for each NLF subject as well as 16 age- and gender-matched controls. To examine contralateral effects, we computed a ratio of the mean pair-wise correlations of all voxels within a ROI with the corresponding contralateral region (IntraRegional Connectivity - IRC). For each subject, ROIs were ranked (from lowest to highest) on ReHo, IRC, and the mean of the two values. At the group level, we observed a significant decrease in the rank for ROI harboring the seizure focus for the ReHo rankings as well as for the mean rank. At the individual level, the seizure focus ReHo rank was within bottom 10% lowest ranked ROIs for all four NLF epilepsy patients and three out of the four for the IRC rankings. However, when the two ranks were combined (averaging across ReHo and IRC ranks and scalars), the seizure focus ROI was either the lowest or second lowest ranked ROI for three out of the four epilepsy subjects. This suggests that rsfMRI may serve as an adjunct pre-surgical tool, facilitating the identification of the seizure focus in focal epilepsy.

Keywords: ReHo; contralateral; epilepsy surgery; focal epilepsy; functional connectivity; non-lesion; pre-operative evaluation; resting state fMRI.

Figure 1

Identification of the seizure focus and ROI. For each NLF subject, the seizure focus was retrospectively defined and identified as tissue showing onset of ictal discharges based on electrocorticographic findings as well as contained with the resection zone (epileptogenic zone). For consistency throughout, each NLF subject is color-coded as seen in Table 1. First column shows a 3D surface rendering of the subjects high-res T1 MPRAGE scan (generated with FREESURFER automated tools for surface reconstruction) with the overlaid ECoG grid and strip electrodes. Electrodes colored in red reveal the locales of the ECoG recorded ictal onset activity. The overlaid red transparencies show the approximate resected, epileptogenic zone. The second column (black boxes) plots the HO ROI (on the MNI 152 brain) that overlaps with the electrode falling within the seizure focus for each NLF subject.

Figure 2

Regions of interest and local connectivity. (A) Shows the 48 thresholded HO ROIs overlaid in the MNI 152 brain. Using the structural detail inherent to the high-res T1 scans, all ROIs were warped into native fMRI space for each subject. Whole brain, normalized ReHo, and IRC values were then extracted and averaged from each ROI. (B) Reveals an example ReHo map form one epilepsy subject. Note the cross-hairs pinpoint a qualitative decrease in ReHo in and around the seizure focus within the right hemisphere, an effect that is absent from the left. (C) Plots raw normalized ReHo values across the 48 HO ROIs for the same NLF subject (green bars) and the mean of the four age- and gender-matched control subjects (white bars). Epilepsy and control values are sorted from lowest to highest for the NLF subject. In this NLF subject, the ROI that contains the epileptogenic zone (ROI 38) has one of the five lowest mean normalized ReHo values of all ROIs.

Figure 3

Ranking the local fc estimates from the HO ROI around the seizure focus. For each NLF subject and the 16 age- and gender-matched controls, ReHo, IRC, and mean scalars were calculated from the ROI that contained the seizure focus. For each of the 20 subjects, values from all ROIs were sorted from lowest to highest and assigned a rank relative to the 48 ROIs within the HO atlas. The first column plots ReHo ranks, the second column plots the IRC ranks (ranking either R > L or L > R) and the third the average rank across the two methods for all subjects. The color bar represents the ranking for the respective NLF subject as noted in Table 1. The gray bar represents the mean (with standard error of the mean) of the 16 controls subjects, and each black bar represents the ranking for each control subject.