Abstracts

Rationale: In healthy human brains, imaging studies have shown that structural connectivity (SC) as well as physical distance constrain but do not fully explain functional connectivity (FC). Beyond estimates of direct SC, search information (SI) and path transitivity (PT), two advanced network metrics that measure ease of information transfer across a network, appear to correlate negatively and positively with resting-state FC, respectively [1]. Here, we investigate whether these structural network metrics can inform information about pathological FC in epilepsy patients, in this case the propagation of interictal epileptiform discharge (IED) activity as observed in intracranial EEG (iEEG) recordings in patients undergoing surgical evaluation.

Methods: 16 epilepsy patients (6 female, ages 32.2 ± 10.6, 5 right hemisphere) were selected who underwent both iEEG monitoring and diffusion magnetic resonance imaging (45 volumes, max b-value 1100, opposite phase encoding). Diffusion images were corrected for motion, eddy current and EPI distortions using TORTOISE, and diffusion tensors and probabilistic tractography were computed in AFNI. Binary SC matrices were computed over 400 whole-brain cortical regions-of-interest (ROIs) defined by a cortical parcellation atlas. From iEEG recordings, IED co-spiking event probabilities between all electrode pairs were computed and thresholded to yield binary FC matrices. SC and FC networks were restricted to ROIs with electrode coverage in the hemisphere of interest. To assess SC-FC similarity, we computed the Dice similarity coefficient between SC and FC (DSC) and the fraction of FC connections with an underlying SC connection (FSC). We then computed the Euclidean distance (D_E), SI and PT averaged over all FC connections in each patient. All 5 metrics were compared to a null distribution computed over 1000 randomly generated, topologically matched networks to determine statistically significant differences. Significance was assessed for the proportion of patients with significant findings for each metric using a chi-squared test (α=0.05).

Results: Of 16 subjects, DSC in 7 and FSC in 5 were significantly greater compared to random SC networks. Compared to random FC networks, mean D_E and SI across FC connections were significantly less in 13 and 11 patients respectively, and mean PT was significantly greater in 3 patients. The number of patients with significant findings was greater than chance for all 5 metrics (Table 1).

Conclusions: IED propagation tends to follow paths with shorter distance and lower search information. Structural networks exhibit significant similarity to IED propagation patterns in a significant but lower number of patients. These results provide evidence that while IED propagation is highly related to distance, diffusion imaging and structural network modeling appear to provide additional information for inferring susceptible pathways of IED propagation.

[1] Goñi J et al. Resting-brain functional connectivity predicted by analytic measures of network communication. Proc Natl Acad Sci 2014;111(2);833-838.

Funding: Please list any funding that was received in support of this abstract.: Funding was provided through the NIH Intramural Research Program.