Abstracts

Rationale: Seizures in temporal lobe epilepsy (TLE) often have cognitive symptoms that require patient self-report for accurate tracking and determining the best therapy. However, this is more difficult in pre-clinical animal models of epilepsy. In this study, we use simultaneous eye-tracking and electrophysiology in an NHP model of TLE to classify gaze behavior between seizure and non-seizure states. We then use a visual paired comparison (VPC) task to further characterize the cognitive effects of TLE seizures in this model.

Methods: A female macaque was head-fixed in front of a blank screen while eye-tracking was used to measure the focal point of each eye with a ~1 kHz sampling rate. 6-13k IU of penicillin-G (PCN) was injected into the hippocampus to induce spontaneous seizures. Local field potentials were recorded from a chronic electrode in the hippocampus at a 2 kHz sampling rate. Seizures were manually annotated and synchronized with the eye-tracking data. The percent time in fixation, percent time in saccade, the x-distance between the left and right eye focal point, and the focal point in the y-direction were extracted. These features were averaged over 2-second windows with a 1-second overlap. Eye-tracking features during seizures were compared to a post-seizure window of the same duration. A support vector machine (SVM) was used to classify between seizure and post. The SVM was cross-validated by holding out each seizure/post for testing while training on the remaining seizures.

A VPC task was used to compare recognition memory between baseline and during seizure sessions. During the task, the subject was presented with an image containing a single object until it focused within a defined area of interest for 30 seconds or 5 minutes had elapsed. The subject’s gaze was then measured when presented with a recollection image composed of the previously seen object and a novel one. A subject with intact memory is expected to spend more time on the novel object. A task was considered invalid if the subject failed to focus on the initial object for fewer than 5 seconds.

Results: For the gaze analysis, 56 seizures were recorded. From seizure to post, there were increases in percent-of-time in saccade (p < 1e-13) and the x-distance between eyes (p < 1e-6), and decreases in the y focal point (p < 1e-9) and the percent-of-time in fixation (p < 1e-20). The SVM had a cross-validated AUC of 0.63±0.03. The VPC task was used to evaluate memory on 200 images at baseline and 179 during seizures. There was no statistical difference in memory performance on the VPC task between baseline and seizures conditions (t-test; p > 0.05). However, there were more invalid tasks during seizure sessions (Fisher exact test; p < 0.05).

Conclusions: We provide evidence showing quantifiable behavioral manifestations of PCN-induced hippocampal seizures. In addition, the results obtained from the VPC task suggest that these seizures do not impair recognition memory – as measured by the task – but may instead affect attention or visual awareness. These findings can provide insight into the cognitive effects of TLE and potential targets for developing new therapies.

Funding: Please list any funding that was received in support of this abstract.: NIH/ORIP P51-OD011132, NIH/NINDS 1UG3NS100559-03.