Quantitative Artifact Reduction and Pharmacological Paralysis Improve Detection of EEG Epileptiform Activity in Critically Ill Patients
Abstract number :
2.007
Submission category :
3. Neurophysiology / 3B. ICU EEG
Year :
2022
Submission ID :
2204591
Source :
www.aesnet.org
Presentation date :
12/4/2022 12:00:00 PM
Published date :
Nov 22, 2022, 05:25 AM
Authors :
Catherine Kulick-Soper, MD – Hospital of the University of Pennsylvania; Russell Shinohara, PhD – Biostatistics, Epidemiology, & Informatics – University of Pennsylvania; Colin Ellis, MD – Neurology – University of Pennsylvania; Taneeta Ganguly, MD – Neurology – University of Pennsylvania; Ramya Raghupathi, MD – Neurology – University of Pennsylvania; Jay Pathmanathan, MD, PhD – Neurology – University of Pennsylvania; Erin Conrad, MD – Neurology – University of Pennsylvania
Rationale: Non-convulsive seizures are common in critically ill patients and carry important treatment implications. Movement-related artifacts, including EMG and myoclonus artifacts, sometimes obscure EEG, limiting detection of epileptiform activity. Both pharmacologic paralysis and quantitative artifact reduction (AR) can reduce EEG artifact. Our goal in this study was to determine the ability of paralysis and quantitative AR to improve the accuracy of detecting epileptiform activity in critically ill patients.
Methods: We performed a retrospective analysis of critically ill patients who underwent continuous EEG monitoring and subsequently underwent pharmacologic paralysis for the purpose of facilitating detection of epileptiform discharges. Four reviewers independently read each patient’s EEG both pre- and post-paralysis, and before and after applying commercially available quantitative AR software (Figure 1). Reviewers indicated the presence or absence of epileptiform discharges. We compared the interrater reliability of EEG interpretation in the baseline, post-AR, and post-paralysis conditions. We compared the performance of AR and paralysis according to artifact type (EMG vs. myoclonus).
Results: We analyzed data from 30 patients (10 women, age range 36-90). The baseline interrater reliability prior to paralysis was only slight (𝜅 = 0.10) with a non-significant trend toward increase after AR (𝜅 = 0.26, p = 0.053 by bootstrapping) and a significant increase after paralysis (𝜅 = 0.51, p = 0.001; Figure 2). When we separated patients according to whether their EEG had high EMG artifact (N = 15), as opposed to primarily myoclonus artifact (N = 15), AR resulted in significant improvement in the group with high EMG artifact but not the group with primarily myoclonus artifact. In the group with high EMG artifact, AR was as effective as paralysis at improving EEG interpretation.
Conclusions: Pharmacologic paralysis improves detection of epileptiform activity in critically ill patients when movement-related artifact obscures EEG features. Quantitative AR improves detection as much as pharmacological paralysis in the setting of high EMG artifact, but AR is ineffective when the primary source of artifact is myoclonus. These results suggest that, when used in the appropriate settings, both quantitative AR and paralysis can facilitate early diagnosis of non-convulsive seizures in critically ill patients.
Funding: E.C. Conrad received funding from NINDS (1K23NS121401-01A1).
Neurophysiology