Career Center AES Connect

Abstracts

Reducing EEG for Fast Seizure Screening

Abstract number : 2.553
Submission category : 3. Neurophysiology / 3A. Video EEG Epilepsy-Monitoring
Year : 2024
Submission ID : 1496
Source : www.aesnet.org
Presentation date : 12/8/2024 12:00:00 AM
Published date :

Authors :
Presenting Author: Rehan Raiyyani, BS – Massachusetts General Hospital

Elizabeth Duquette, MD, BS – University of Michigan
Hannah Culbertson, BS – University of Utah
Mitchell Couldwell, MBA, MS, BS – Tulane University School of Medicine
Andrew Zayachkivsky, PhD – University of Utah
F. Dudek, PhD – University of Utah
Kevin Staley, MD – Harvard Medical School

Rationale:

In order to find spontaneous seizures to determine the presence of epilepsy in longitudinal electroencephalography (EEG), animals are monitored for many months. This generates hundreds of days of data, making manual reading difficult. Thus, strategies to reduce workload are critical to analyze data from long-term seizure-monitoring experiments. Here, we propose two strategies that may substantially reduce the length of EEG for manual reading: (1) reducing the quantity of EEG by initially detecting seizure candidates (White et al. 2006), concatenating detected segments, and then removing epochs not found to have any seizure(s) with a custom script and (2) removing EEG epochs less likely to contain seizures by analyzing their power spectra (Donoghue et al. 2020). To evaluate the feasibility of these approaches, we compared the durations and seizure quantity found using the first data reduction method, and after applying both data reduction methods.



Methods:

Seven rat subjects were selected that had either undergone hypoxia-ischemia (HI) or were in the sham surgical control group. Animals in HI group may not have developed stroke, but had acute seizures and various other surgery- and implantation-related injuries. Animals were continuously monitored for development of epilepsy using 2 channel EEG wireless telemetry. Subjects were screened for seizures after modifying the EEG recording with 2 different approaches to reduce the data. Method I identified potential ictal events using a detection software (dClamp), merging events, and then removing EEG epochs that did not contain seizure candidates. Method II added power spectral processing to the first method to remove additional EEG epochs. Candidate ictal onsets marked (using Method I and Methods I + II) were then manually reviewed using raw recording traces.



Results:

The initial EEG duration summed across 7 animals was 107 days; the first method reduced the cumulative EEG data duration to 15 days; adding the second method to the first reduced the cumulative duration to 3 days (with all 3 category durations rounded to nearest day respectively). To date, 33 seizures were identified and pooled from 3 of the 7 subjects when using the first data reduction method while 26 seizures were identified (and pooled from the same 3 of the 7 subjects) when combining the second and the first methods. The quantity of electrographic seizures identified was 29 vs. 16, 4 vs. 9, and 0 vs 1 (for method I vs. method I + method II approaches respectively) for the 3 rats with seizures.



Conclusions:

While data reduction accelerates manual seizure screening, the reduced dataset may still be too large to review manually without substantial human error. False positives and false negatives can be found in many published studies which can lead to misinterpretation. Adding a second data reduction method further reduced the EEG but resulted in >= 13 missed seizures (false negatives). We are continuing to improve data reduction method(s) to retain all ictal events and minimize the amount of EEG to be manually reviewed.



Funding:

Supported by NINDS 5R01NS086364 and DOD-CURE contract W81XWH-15-2-0069.



Conflicts of Interest:
F.E. Dudek has had equity interest in and has received remuneration from Epitel, Inc. and Cage Data Corp. for work related to wireless EEG devices for rodents and humans. A. Zayachkivsky has received consulting fees from Cage Data Corp. and BIOPAC Systems for work on the EPOCH telemetry device.


Neurophysiology