Authors :
Presenting Author: Tonmoy Monsoor, PhD – UCLA
Sotaro Kanai, MD, PhD – Division of Pediatric Neurology, Department of Pediatrics, David Geffen School of Medicine at the University of California, Los Angeles, California, USA
Atsuro Daida, MD, PhD – Saitama Children's Medical Center
Prateik Sinha, BS – UCLA
Naoto Kuroda, MD, PhD – Wayne State University
Shingo Oana, MD, PhD – UCLA
Lawrence Liu, BS – UCLA
Gaurav Singh, MS – UCLA
Yipeng Zhang, MS, PhD – Department of Electrical and Computer Engineering, University of California Los Angeles
Chenda Duan, MS – UCLA
Lina Zhang, MS – School of Engineering, UCLA
Shaun A. Hussain, MD, MS – Division of Pediatric Neurology, Department of Pediatrics, UCLA Mattel Children's Hospital, David Geffen School of Medicine
Raman Sankar, MD, PhD – UCLA
Aria Fallah, MD, MSc, MBA – Department of Neurosurgery, UCLA Medical Center, David Geffen School of Medicine
Richard J. Staba, PhD – Department of Neurology, UCLA Medical Center, David Geffen School of Medicine
Jerome Engel Jr., MD, PhD – Department of Neurology, UCLA Medical Center, David Geffen School of Medicine
Eishi Asano, MD, PhD – Wayne State University
Vwani Roychowdhury, PhD – Department of Electrical and Computer Engineering, University of California Los Angeles
Hiroki Nariai, MD, PhD, MS – Department of Pediatrics, Division of Pediatric Neurology, David Geffen School of Medicine at the University of California, Los Angeles, California, USA
Rationale:
Identifying the epileptogenic zone (EZ) in patients with medication-resistant epilepsy using brief interictal intracranial EEG (iEEG) remains a major clinical challenge.Current practice involves iEEG monitoring-often several weeks-to capture seizures and define the seizure onset zone (SOZ).However,this approach often fails to delineate the entire EZ and yields successful surgical outcomes in only 60-70% of cases.We hypothesize that epileptic network dynamics form a continuum,with the same neurophysiological mechanisms active during both ictal and interictal periods.Specifically,we propose that the EZ actively drives spontaneous interictal hypersynchronous network (HSN) events,similar to it’s role in seizure generation.Thus,detecting and characterizing interictal HSNs may enable accurate delineation of the EZ.
Methods:
We studied 159 pediatric patients at UCLA and Wayne State University who underwent chronic iEEG monitoring with grid/strip and subsequent resection.All patients had SOZ annotations and postoperative seizure outcomes assessed after one year. For any given iEEG recording segment (sampled at 1,000 Hz), we computed directed and weighted synchronization coefficients between pairs of channels using power-phase coupling in high-frequency bands(gamma:50-80Hz,ripple:80-250Hz,and fast ripple:250-300Hz) to construct synchronization networks among channels. This resulted in a sequence of networks from a 5-minute interictal iEEG; one set for each segment.For each channel,we computed a feature vector leveraging the temporal dynamics of its network properties,which was used to train and test an unsupervised machine learning model to estimate the likelihood of a channel belonging to the EZ (EZ centrality score).Lastly,we built a random forest model incorporating the resection status of estimated EZ to predict postoperative seizure freedom.
Results:
We found intermittent emergence of interictal HSNs (mini-seizures) where (i) some of the interictal drivers are also labeled as SOZs and appear as drivers in the ictal HSNs (seizures),and (ii) there are other interictal drivers with the same network characteristics as the SOZs(Fig 1).The unsupervised class (drivers vs non-drivers) discovery and subsequent labeling framework identified the drivers of interictal HSNs,which we predicted as comprising the EZ(Fig 2(A)),from interictal iEEG with duration as short as 3 minutes(Fig 2(B)).Our postoperative seizure outcome prediction model based on the resection status of the predicted EZ yielded a mean F1 score of 87%(p< 0.001),which was significantly higher than the predictive power of current practice utilizing the resection status of the SOZ(F1 score-78%) and HFO with spikes(spkHFO) (F1 score-79%)(