Causal Network Mapping of sEEG Identifies Compact Epileptogenic Targets Concordant with Seizure Freedom: Multicenter Validation in 60 Patients
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background and Purpose
Drug-resistant epilepsy (DRE) affects approximately 15 million people worldwide, and surgery remains one of the only curative options. A key challenge in predicting outcomes is the lack of standardized, quantitative tools to help distinguish seizure “driver” regions from “responder” regions during stereoelectroencephalography (sEEG) recordings. We validated a novel metric we call criticality which uses causal network mapping and machine learning to assign scores to sEEG contacts such that higher scores correspond to surgically treated tissue in patients with more favorable outcomes. Criticality is a per-contact score of how strongly each recording site drives the seizure network. Compared to the most similar previously published method, neural fragility, criticality also measures which nodes in the network have destabilizing (seizure-causing) network effects but also models time-delayed connections and frequency-specific interactions such as phase-amplitude and cross-frequency coupling.
Methods
We analyzed de-identified clinical data from 60 patients (aged ≥ 2 years) with focal or multifocal DRE who underwent sEEG monitoring and proceeded to surgery at four U.S. Level-4 epilepsy centers. The algorithm was trained on an independent cohort (N=37) and locked prior to validation. A random forest mapped each patient’s distribution of criticality values inside versus outside the treatment zone (TZ) to a probability of surgical success (P(success)), evaluated by leave-one-patient-out cross-validation. The primary outcome was the standardized effect size (Cohen’s d) of P(success) on the validation cohort between more favorable (Engel I–II) and less favorable (Engel III–IV) patient outcomes.
Results
The predicted P(success) was significantly higher in patients with more favorable outcomes in our held-out validation dataset (d = 1.12, 95% CI: 0.42–1.83, p = 0.001). Three potentially clinically actionable findings emerged: 1) High-criticality contacts formed spatially compact clusters (∼9 mm nearest-neighbor distance vs. 17 mm expected by chance), consistent with focal targets amenable to minimally invasive ablation. 2) Sensitivity was highest in small focal procedures (80% at ≤10 treated contacts) and decreased with resection size. 3) In patients with less favorable outcomes, high-criticality tissue remained outside the resection boundary, suggesting incomplete resection of the epileptogenic zone.
Conclusions
Our criticality metric applied causal network mapping to sEEG recordings, improving on the state-of-the-art benchmark for predicting treatment success probability in retrospective cases. Our criticality metric performed best in focal procedures and may be best suited for laser interstitial thermal therapy (LITT) and other minimally invasive approaches. When seizures persisted after surgery, residual high-criticality tissue outside the resection boundary offered both a mechanistic explanation for the less favorable outcome as well as potential targets for reoperation.
