Abstracts

Rationale: Multiple studies have reported that the majority of seizures occurring in critically ill patients are nonconvulsive and can be recognized only by continuous EEG (cEEG) monitoring. For many patients, early diagnosis and treatment of nonconvulsive seizures (NCSs) may prevent pathophysiologic effects of seizures caused by a primary insult to the brain. Due to advances in computer and networking technology, the use of cEEG monitoring in ICU patients has expanded rapidly. However, fewer than 10% of ICUs currently offer cEEG monitoring. One factor contributing to this low rate of ICU cEEG monitoring is the expertise and resources required to reliably interpret these cEEG recordings. A reliable algorithm for EEG seizure detection in the ICU could encourage expanded use of ICU cEEG monitoring. We report a preliminary investigation of a novel automated algorithm for the detection of NCSs in cEEG recordings from ICU patients.Methods: Twenty-four prolonged cEEG recordings obtained from 22 ICU patients were studied (total duration ~ 1,500 hours, with a total of 453 NCSs in 11 recordings). Datasets were not pre-selected for analysis and therefore included ICU patients with a variety of etiologies and ictal EEG patterns. Using quantitative EEG (qEEG) measurements for signal regularity, power, and maximal frequency (all calculated sequentially for each nonoverlapping 5.12 sec epoch), the proposed detection algorithm detects significant changes (i.e., an increase of signal regularity and/or power, as seen in most ictal EEGs) of qEEGs from the preceding baseline means as well as the asymmetry between hemispheres. The algorithm also utilizes signal power and frequency to reject artifacts such as recording noise, muscle/movement, and electrode failure. For each cEEG recording, the algorithm generated a list of detection times through the analysis of the patterns in qEEG measures. Based on the clinical seizure reports, we evaluated the performance of the algorithm by estimating the detection sensitivity and the false detection rate (per hour). Using the same datasets, we also assessed the detection performance of two commercial seizure detection products, which were clinically validated with only cEEG recordings from epilepsy monitoring units.Results: Overall, our proposed algorithm accurately detected 405 of 453 NCSs (mean sensitivity = 90.4%) with a mean false detection rate of 0.066/h (or 1.59/day). The two commercial detection products performed at low sensitivities (mean = 12.9% and 10.1%) with false detection rates of 1.036/hr (or 24.9/day) and 0.013/h (0.30/day), respectively.Conclusions: These findings suggest that, with further optimization and clinical validation, the proposed seizure detection algorithm, specifically designed for detecting NCSs in ICU patients, has potential to be the basis of clinically useful software that can assist ICU staff in timely identification of NCSs. This study also suggests that currently available seizure detection software does not have sufficient performance for the detection of NCSs in critically ill patients.