Deep Learning Classifier for the Combined Scoring of Sleep-wake and Seizures in Mice
Abstract number :
3.2
Submission category :
3. Neurophysiology / 3G. Computational Analysis & Modeling of EEG
Year :
2022
Submission ID :
2205041
Source :
www.aesnet.org
Presentation date :
12/5/2022 12:00:00 PM
Published date :
Nov 22, 2022, 05:27 AM
Authors :
Brandon Harvey, BS – Emory University / University of California Davis; Lauren Aiani, BS – Lab Manager, Neurology, Emory University; Viktor Oláh, BS – Emory University; Lucie Rosenberg, BS – Research Specialist, Neurology, Emory University; Matthew Rowan, Ph.D. – Assistant Professor, Cell Biology, Emory University; Nigel Pedersen, M,D. – Assistant Professor, Neurology, Emory University
Rationale: The relationship between sleep and seizure is complex and bidirectional, including seizure-frequency-associated sleep fragmentation as well as sleep-deprivation-induced increases in seizures. In order to provide better throughput for sleep studies in Intra-Amygdalar Kainic Acid (IAKA) Temporal Lobe Epilepsy (TLE) model mice, we have designed a data processing pipeline and Keras-based deep learning classifier for sleep staging and seizure scoring. While sleep-wake classifiers and seizure detection algorithms abound, the combination is difficult given the abnormal EEG background in mice with epilepsy.
Methods: We obtained electrocorticogram (ECoG), electromyogram (EMG), video and bilateral hippocampal depth electrode data from mice implanted using a customized 3D-printed headplate designed in our lab. Our dataset included 1300 12-hour recordings from a group of 47 mice (including 23 mice that developed spontaneous seizures and 10 control mice), that included 642 seizures. We used 70% of the data for training and 30% for testing. The classifier was trained on manually scored 20-second epochs from existing downsampled recordings (from 2kHz to 200Hz), using Fourier bins of alpha, beta, delta, gamma, and theta frequency bands in recordings from this montage.
Results: With a sequential model, the classifier has achieved >90% scoring accuracy in most categories for our epileptic mice. This reliable classification will allow for rapid combined sleep-wake and seizure scoring in our IAKA sleep-wake paradigm.
Conclusions: Using this classifier to improve our throughput, we hope to improve the mechanistic and symptomatic understanding of sleep disruption in epilepsy, leading to improved patient treatment and outcomes.
Funding: CURE Grant
Neurophysiology