Abstracts

Rationale: Rodent models of clinical epilepsy are indispensable for probing disease circuits and testing novel therapies. Several models have been created in order to mimic specific aspects of human epilepsy. These models provide a controlled experimental framework in which to test hypotheses concerning seizure initiation, buildup, spread, and termination. However, to effectively use these models, an epilepsy monitoring unit (EMU) must be designed to be capable of performing high-quality extended continuous recordings, coupled to an efficient and robust analytical pipeline. The EMU should be enabled with both on-line computational processing and off-line parameter extraction, event detection, and detailed semi-automatic annotation. We have developed a set of hardware and software tools to acquire and analyze video-EEG data, including a detector for seizures, interictal spikes, and other events. Our data is also easily annotated to provide labels for future classification studies. With this acquisition and analysis pipeline, our rodent EMU presents a tool for generating and processing large amounts of data from rodents with epilepsy. We envision this approach to video-EEG data being applied to a variety of diagnostic and therapeutic studies, including in the clinical EMU setting. Methods: We implanted young (aged 2-3 months, n = 12) rats with surface electrodes, EMG pads, and intrahippocampal depth electrodes bilaterally. We subsequently injected these rats intrahippocampally with chemoconvulsants: kainic acid (n = 4), tetanus toxin (n = 4), and pilocarpine (n = 4). We then conducted video-EEG recordings continuously for 3 months using custom MATLAB software. We analyzed the EEG data by computing features in the time, frequency, and synchronization domains. We identified events of interest including seizures and interictal spikes by thresholding select features. We then utilized PCA and clustering approaches to identify additional classes of events. Results: We successfully recorded video-EEG data from our rat models of focal epilepsy. We characterized the electrographic recordings in terms of band power, entropies, Hjorth parameters, and other features. In one tetanus-injected rat, we describe 22 Racine 4-5 seizures (duration 36.0 s  ± 9.6 s) over the course of 13 days. In another kainate-injected rat, we describe nine seizures (duration 55.6 s ± 16.7 s) with varied Racine staging over the course of 70 days. In addition to seizures, we comprehensively identified inter-ictal spikes, and spike-wave discharges. Conclusions: We have developed a system in which it is possible to acquire and process video-EEG continuously and efficiently from rodent models of epilepsy. The big data approach used in this study allows for easy application of machine learning algorithms for brain state classification. This rodent EMU has been employed successfully to characterize electrographic and behavioral events in these models. Funding: NIH T32MH020017-19NIH T32GM007753-37Paul & Daisy Soros Fellowship