Abstracts

Rationale: Defining the network(s) of seizure initiation and propagation is paramount for our understanding of epilepsy, epileptogenesis, and the development of new treatment modalities. To understand and explore the “network theory” in epilepsy, we used a novel animal model of mesial temporal lobe epilepsy (MSO model) implanted with multiple cortical and hippocampal tetrodes. Here, we highlight a pipeline to automate the detection of epileptic events and exploration of networks in this model.

Methods: Male Sprague-Dawley rats were surgically implanted with a microdialysis cannula and osmotic pump infusing methionine sulfoximine (MSO), a glutamine synthetase enzyme inhibitor targeting the ventral dentate gyrus and a chronic microdrive containing multiple, independently movable tetrodes/electrodes in a separate craniotomy targeting the CA layers. After recovery from the surgery, the tetrodes were lowered into cortex and hippocampal CA1, CA3 and dentate gyrus. Signals from each channel were acquired using a high-density Digital Lynx data acquisition system (Neuralynx, Bozeman, MT). Each animal then underwent recording sessions in which the animal was placed into a large home pot and connected to the acquisition system with video monitoring for several hours. Similar to seizure monitoring in the EMU, we sought to automatically identify spontaneous epileptic events from each animal. We used Teager energy’s operator to identify and detect epileptic discharges or seizures. Three expert reviewers independently reviewed the events detected by our algorithm and score them as Polyspikes, Brief Ictal Rhythmic Discharge (BIRD), Seizure, Noise, and Unclear.

Results: We used 2 separate recordings from 3 animals (n = 6) with average length of 119.1 minutes (±36.9SD). In total, the pipeline detected 135 events from these recordings. Cumulative results of three independent reviewers were: Polyspikes (58.4%), BIRD (21.3%), seizure (6.84%), noise (7.52%), and unclear (5.81%). Therefore, in a short time, this pipeline detected epileptiform discharges with an error rate of 13.46%.

Conclusions: These preliminary results highlight a potentially robust pipeline that is capable of automatic detection of epileptic discharges and seizures. The event detection system can further guide us to the periods of epileptogenesis across time and brain regions.

Funding: Please list any funding that was received in support of this abstract.: Yale Center for Clinical Investigation (YCCI) Scholar Award & Swebilius Foundation.