Abstracts

Rationale: Up to 53% of patients with traumatic brain injury (TBI) develop spontaneous recurrent epileptic seizures. The ability to predict epileptogenesis in these patients may reduce the burden of post-traumatic epilepsy by enabling earlier interventions in the disease process as well as personalized therapy according to the likelihood of developing epilepsy. Here, we use a two-stage machine learning detection algorithm and analyze the utility of postulated epileptiform spikes and bursts as potential electrographic biomarkers of epileptogenesis in a rat model of TBI.Methods: Hippocampal depth recordings from 23 rats monitored continuously for a one-week period three months following a fluid-percussion-induced traumatic brain injury (17) or sham surgery (6). All recordings were obtained prior to first seizure. Eight of the 17 rats with injury developed spontaneous seizures after the initial recording period. Candidate bursts were initially detected with a line-length filter and both unsupervised and supervised methods were used to cluster epileptiform spikes and bursts into artifact and non-artifact groups. Permutation tests across rats were used to assess statistical significance in spikes and bursts across three groups (sham, no-seizure, and seizure).Results: Epileptiform spikes and bursts were detected in all groups that received surgery and electrode implantation. Three subtypes of bursts were detected and subsequently classified into rhythmic bursts, epileptiform bursts, and electrode artifact. There was no observed difference in the number of spikes between the sham, seizure, and non-seizure groups. Among the 17 rats that received injury, those that developed seizures displayed more epileptiform bursts per day than the rats that did not develop seizures, though this did not reach significance in the given sample.Conclusions: Our results show that in this sample of 17 rats, there is a trend that epileptiform bursts are more prevalent in rats that develop seizures versus rats that do not. There was no difference in interictal spikes between groups. This work encourages further investigation of bursts as a potential biomarker of epileptogenesis.