Method for Localizing the Seizure Onset Zone in Refractory Epilepsy Patients

In recent years, refractory epilepsy has received increasing attention from the medical community. Refractory epilepsy is defined as the continuing occurrence of severe seizures despite treatment with two appropriate antiepileptic drugs. For patients who are unresponsive to drug treatment, if the seizure onset zone (SOZ) can be accurately localized, treatment methods such as resection or ablation of that area may have a curative effect. However, in the United States, monitoring epileptic activity in different brain regions using stereoelectroencephalography (SEEG) electrodes is a common surgical evaluation method for patients with drug-resistant epilepsy, but this method relies on detecting a sufficient number of seizures, requiring patients to be hospitalized for monitoring for several days or even weeks. Furthermore, even after completing SEEG monitoring, precise localization of the SOZ cannot be guaranteed. Therefore, research on how to improve the accuracy of SOZ localization is of significant importance.

The main authors of this paper include Alexander G. Yearley, Elliot H. Smith, Tyler S. Davis, Daria Anderson, Amir M. Arain, and John D. Rolston, from Harvard Medical School, Brigham and Women’s Hospital, University of Utah, and the University of Sydney. The paper was published in the Journal of Neurosurgery with an online publication date of May 24, 2024. The research aimed to develop a computational approach to identify and characterize interictal epileptiform discharges (IEDs) from SEEG electrode recordings and to localize the SOZ using the propagation directionality of IEDs.

Research Methods

Research Participants

The study recorded SEEG from 15 patients with drug-resistant epilepsy at the University of Utah between 2019 and 2021. To be included in the study, patients needed at least 16 hours of SEEG recording and corresponding tractography imaging. Patients underwent neuromonitoring for 2 to 8 days to localize the epileptic focus. The electrode implantation was based on pre-SEEG hypotheses, determined through a multidisciplinary conference. The presumed SOZ was concentrated in certain specific areas, with electrodes implanted in those areas based on prior electroencephalography, preoperative imaging, seizure semiology, and neuropsychological findings. Additionally, for patients with lesions, if the lesion was considered a potential source of seizures, the surrounding cortex was sampled as a potential source region.

IED Detection and Propagation Wave Quantification

The study detected IEDs from continuous SEEG recordings of the 15 patients and identified IEDs using a modified peak detection algorithm. Each IED required a peak detection in a 6-millisecond window across at least 5 channels to be marked as an IED. To avoid detecting incidental signal noise, IED detection had to meet the definition criteria set by the International Federation of Clinical Neurophysiology.

To determine the nature of IED propagation waves, the study used three measurement methods: Euclidean distance between electrodes, tractography-determined axonal path length, and connectivity probability between electrodes. By calculating IED propagation velocity, the study further quantified whether IEDs met the criteria for a propagation wave.

Spatial Relationship Between IEDs and SOZ

The SOZ was defined by clinical neurologists based on epileptic activity in the continuous SEEG recordings. To evaluate the extent of the SOZ, the study performed a sensitivity analysis using four different definitions of the SOZ: 0 mm, 10 mm, 20 mm, and 30 mm. The propagation paths of IEDs were determined by the time sequence of SEEG electrode contacts, with each IED detection parsed into functional units called “triplets,” each consisting of three consecutive contacts that detected the same IED. By analyzing the frequency and distribution of triplets, the study investigated the proportion and pathways of IEDs through the SOZ.

Research Results

The study found that an average of 23.2 hours of SEEG data was recorded per patient, with a median (range) of 22.6 (4.4-183.9) IEDs detected per hour. On average, 61.8% of IEDs exhibited propagation wave characteristics, with a median IED propagation wave velocity of 44.5 cm/s based on Euclidean distance and 126.1 cm/s based on axonal path length. For different definitions of the SOZ location, the study found significant differences in the proportion of IEDs passing through the SOZ: with the narrowest definition (0 mm), the median proportion of IEDs originating from the SOZ across patients was 17.4%, while the median proportion of overall pathways through the SOZ was 20.8%. With the broadest definition (30 mm), the proportion of IEDs originating from the SOZ increased to 62.1%.

High-frequency triplets were more likely to propagate through the SOZ, with 100% of high-frequency triplets appearing in at least one IED that passed through the SOZ. Further regression analysis showed that patients with more IEDs detected per hour, a higher proportion of IED propagation waves, and higher triplet frequency were more likely to have IEDs localized to the SOZ.

Conclusions and Significance

Through computational methods, IEDs can be effectively detected from clinical SEEG recordings of epilepsy patients and associated with preliminary definitions of the SOZ. While IEDs exhibited heterogeneity in SOZ localization across different patients, patients with a higher proportion of IED propagation waves had IEDs significantly more likely to localize to the SOZ. Triplet analysis of IEDs helped distinguish stable pathways through the SOZ from non-passing pathways, and these findings provide potential biomarkers for SOZ mapping.

The study provides a possibility for introducing IED analysis into the clinical workflow, potentially positively impacting this high-risk patient population by reducing the monitoring time and the number of implanted electrodes required to delineate the seizure onset region. Future work needs to validate these methods in larger populations and investigate the relationship between IED features and clinical outcomes. This study supports the potential application of detecting IEDs and their propagation waves in SOZ localization for epilepsy patients.