Authors :
zhinoos Razavi, PhD – Melbourne University; Mark Cook, Professor – Melbourne University; Levin Kuhlmann, PhD – Monash University
Rationale: Epilepsy is a neurological disorder that is diagnosed by occurrence of seizures more than once. Duration of seizures are varied and plays a key role in the severity and associated risks for epileptic patients. Newly, Liu et al proposed a seizure duration prediction algorithm that categorize duration of a seizure into short or long from its onset. They tested their proposed algorithm on full iEEG data record of the 15 patients from the long-term trial of the Melbourne-NeuroVista seizure trial and the
Melbourne-University AES-MathWorks-NIH Seizure Prediction Challenge. They reported the AUC performance of 0.7 for 5 patients out of 10 patients. Inspired by the proposed seizure duration algorithm, we investigated circadian and multidien pattern of seizure duration to predict status epilepticus and optimize battery life of seizure control devices.
Methods: We extracted ICTALS of 15 individual patients with their duration from the full trail of NeuroVista data. All ICTALS were classified as short or long based on a ration distribution threshold.
Results: Statistical analysis of circadian distribution of 3319 short and long seizure showed seizure occurrences of 10 out of 15 patients are more frequent during the night (9:00 PM-9:00 AM) than the day (9:00 AM-9:00 PM) with the average ratio of %64.40. It also demonstrated that long seizures are more prone to happen during the night with the average ratio of %64.28. Monthly analysis of 13 out of 15 patients over period of 3 years showed the number of long seizures during the night is more frequent than during the day.
Conclusions: Current results suggest that the amount of electrical stimulation delivered by seizure control devices can be optimised by considering circadian and multidien seizure duration.
Funding: Funding: NHMRC GNT1160815