Authors :
Matthew Pesce, BS – Boston Children's Hospital; Robert Billardello, MSc – Research Unit of Advanced Robotics and Human-Centred Technologies, – University Bio-Medico Campus of Rome; Jeffrey Bolton, MD – Neurology – Boston Children's Hospital; Joseph Madsen, MD – Department of Neurosurgery – Boston Children's Hospital; Phillip Pearl, MD – Neurology – Boston Children's Hospital; Christos Papadelis, PhD – Jane and John Justin Neurosciences Center – Cook Children’s Health Care System; P. Ellen Grant, MD – Fetal-Neonatal Neuroimaging and Developmental Science Center – Boston Children's Hospital; Eleonora Tamilia, PhD – Fetal-Neonatal Neuroimaging and Developmental Science Center – Boston Childrens Hospital
This abstract is a recipient of the Young Investigator AwardRationale: For children with drug-resistant epilepsy (DRE), epilepsy surgery aims to remove the epileptogenic zone (EZ). The gold standard for estimating EZ is localizing the seizure onset zone (SOZ) by recording seizures with intracranial EEG (iEEG), which yet may require many days of recordings. Estimating the EZ via interictal data – with no need for seizures – would largely benefit presurgical planning, but an undisputable interictal biomarker of the EZ does not exist. A traditional interictal approach involves delineating the irritative-zone (IZ: iEEG contacts showing spikes). Yet, IZ lacks specificity and is a local biomarker, while epilepsy is now conceptualized as a brain network disorder, where Functional Connectivity (FC) analysis can identify hyperconnected seizure networks. Based on this premise, we present a novel iEEG interictal measure to estimate the EZ introducing the concept of irritative-network: we integrate traditional spike identification on iEEG (IZ) with FC estimation between contacts (functional-network). We aim to demonstrate that our novel irritative-network index improves upon the traditional IZ at estimating the EZ.
We studied iEEG (5-min, Figure 1A) of epilepsy surgery patients from Boston Children’s Hospital with known outcomes (Engel). For each patient, we: _x000D_
_x000D_
Detected spikes and computed spike-count per iEEG contact - Figure 1B;_x000D_
Performed FC analysis, obtaining one FC-matrix per patient - Figure 1C;_x000D_
Estimated a new irritative-network index (Figure 1D; product of connectivity-matrix and spike-count) that quantifies the spiking rate of each contact plus the spiking rates of the connected contacts._x000D_
_x000D_
To assess the presurgical value of our irritative-network index, we compared (Wilcoxon sign-rank test) values inside and outside the SOZ (defined by reports) or resection (defined by postop-MRI), and computed their ROC curves and area-under-the-curve (AUC) to quantify the ability to identify SOZ and resection. We identified the top irritative-network contacts per patient by thresholding, measured their overlap with resection and assessed whether the irritative-network index could predict outcome (AUC; Fisher’s test).
Results: We analyzed data from 40 children (26 Engel1; age:12±5 years). The Irritative-network index (Figure 2A) is higher inside than outside the SOZ and identifies the SOZ better than traditional spike-count (AUC: 0.78 vs 0.76, Figure 2B). Regarding the resection, the Irritative-network index (Figure 2C) is higher inside than outside and outperforms spike-count in identifying the resected contacts in good-outcomes (AUC=0.68 vs 0.56, Figure 2D). Finally, removal of the irritative-network contacts outdoes spike-count (AUC=0.74 vs 0.63) in predicting surgical outcome (Figure 2E) with positive and negative predictive value of 64% and 71% (p=0.024).
Conclusions: The irritative-network concept outperforms the traditional IZ in estimating the SOZ and the area to resect during surgery in children with DRE. Our findings suggest that integrating traditional IZ concept with FC analysis offers an additional presurgical tool to estimate the EZ without needing to capture seizures, thus reducing recording time.
Funding: BCH/Harvard Faculty Career Development Fellowship