CRCNS: Following the BOLD lightening at rest strikes the seizure onset zone!

Epilepsy is a devastating disease affecting over 50 million people worldwide (WHO). About 30% of patients do
not respond positively to medication and are diagnosed as having drug resistant epilepsy (DRE). DRE causes
significant costs, morbidity, and mortality. The most effective treatment is to surgically remove the seizure onset
zone (SOZ), the region from which seizure activity is triggered. The localization of the SOZ is essential for
surgical success. Unfortunately, surgical success rates range from 30%-70% because there is no reliable
biomarker of the SOZ. We propose to develop a combined intracranial EEG-fMRI biomarker of the SOZ while
the patient is not seizing or at “rest”. One may ask, “how does one identify where seizures start in the brain
without ever observing a seizure, and if this is possible why have previous methods failed?” The fundamental
limitation of current computational approaches for both resting state fMRI (rs-fMRI) and intracranial EEG (rsiEEG)
SOZ localization lies in the fact that they compute static measures from observations produced by a
dynamic epileptic network. We believe that a computational method that can provide a characterization of how
the observations are dynamically generated in the first place, and how internal network properties can trigger
seizures or prevent seizures will be successful in SOZ localization. Therefore, we will construct dynamical
network models (DNMs) in this study. DNMs are generative models that capture how every network node
(location of centralized network signal processing and transfer) interacts with every other node dynamically.
DNMs uncover internal properties including bandwidth, stability, controllability, system gain, and most important
to this application - connectivity. We propose that when a patient is not having a seizure, it is because the SOZ
is being inhibited by neighboring nodes (brain regions). We thus will apply DNM algorithms in a novel manner to
identify two groups of network nodes from rs-fMRI and rs-iEEG: those that are continuously inhibiting a set of
their neighboring nodes (denoted as “sources”) and the inhibited nodes themselves (denoted as “sinks”). Thus,
in line with the most recent advancement in precision medicine, for each patient, we will build DNMs customized
to identify and quantify, via a score, key sources and sinks, optimized to localize the primary causative SOZ
nodes in the epileptogenic network and their connectivity properties. We will leverage functional imaging data
while patients are “at rest” in a study population of children with DRE who are undergoing epilepsy surgery
evaluation. Specifically, we will construct DNMs from rs-fMRI and rs-iEEG data and test our novel “source-sink”
hypothesis that may point to the SOZ when patients are not seizing. If successful, the proposed DNMs could
significantly increase surgical candidacy and improve surgical outcomes by increasing the yield of surgically
actionable results and precision of SOZ localization. Furthermore, by removing the need to capture seizures,
this novel dynamic network model-based SOZ localization biomarker may substantially reduce invasive
monitoring times, avoiding further risks to patients and reducing costs to hospitals.

Grant Number: 5R01NS125897-05
NIH Institute/Center: NIH
Principal Investigator: Varina Boerwinkle