Noninvasive Epileptogenic Signal Direction Determination by Effective Connectivity of Resting State Functional MRI - Independent of EEG

Abstract

The goal of this study was to determine resting state fMRI effective connectivity (RSEC) capacity, agnostic of epileptogenic events, in distinguishing seizure onset zones (SOZ) from propagation zones (pZ). Consecutive patients (2.1-18.2 years old), with epilepsy and hypothalamic hamartoma, pre-operative rs-fMRI-directed surgery, post-operative imaging, and Engel class I outcomes were collected. Cross-spectral dynamic causal modelling (DCM) was used to estimate RSEC between the ablated rs-fMRI-SOZ to its region of highest connectivity outside the HH, defined as the propagation zone (pZ). RSEC between the SOZ and PZ was characterized as positive (excitatory), negative (inhibitory), or null. It was hypothesized that connectivity from the SOZ would be excitatory and connectivity from the pZ would be inhibitory. Sensitivity, accuracy, positive predictive value were determined for node-to-node connections. A Parametric Empirical Bayes (PEB) group analysis was performed to identify effects of Engel class outcome and age. RSEC strength was also evaluated for correlation with percent seizure frequency improvement, sex, and region of interest size. Of the SOZ's RSEC, only 3.6% had no connection of significance to the pZ when patient models were individually reduced. Among remaining, 96% were in expected (excitatory signal found from SOZ→pZ and inhibitory signal found from pZ→SOZ) versus 3.6% reversed polarities. Both polarity signals were equivalently as expected, with one false signal direction out of 26 each (3.7% total). Sensitivity of 96%, accuracy of 93%, and positive predictive value of 96% in identifying and differentiating the SOZ and pZ. Groupwise PEB analysis confirmed SOZ→pZ EC was excitatory, and pZ→SOZ EC was inhibitory. Patients with better outcomes (Engel Ia vs. Ib) showed stronger inhibitory signal (pZ→SOZ). Age was negatively associated with absolute RSEC bidirectionally, but had no relationship with Directionality SOZ identification performance. This study demonstrates the accuracy of Directionality to identify the origin of excitatory and inhibitory signal between the surgically confirmed SOZ and the region of hypothesized propagation zone in children with DRE due to a HH. Thus, this method validation study in a homogenous DRE population may have potential in narrowing the SOZ-candidates for epileptogenicity in other DRE populations and utility in other neurological disorders.