Individualized Neural Dynamics Patterns During Sleep by High-Resolution Phenotype Mapping
Poster No:
2587
Submission Type:
Abstract Submission
Authors:
Hyeon Jin Kim1,2, Younghan Lee3, Sophie Saremsky4, Habiba Noamany4, Youngseo Kim5, Ho Bae5, Yunheung Paek3, Michael J. Prerau4, Hyang Woon Lee5,2
Institutions:
1Department of Neurology, Korea University Ansan Hospital, Ansan, Republic of Korea, 2Departments of Neurology and Medical Science, Ewha Womans University School of Medicine and Ewha Medical Research Institute, Seoul, Korea, Republic of, 3Department of Electrical and Computer Engineering (ECE) and Inter-University Semiconductor Research Center (ISRC), Seoul National University, Seoul, Korea, Republic of, 4Division of Sleep and Circadian Disorders, Brigham and Women’s Hospital, Boston, MA, 5Artificial Intelligence Convergence and Computational Medicine, System Health Science & Engineering, Graduate School of Ewha Womans University, Seoul, Korea, Republic of
First Author:
Hyeon Jin Kim, MD
Department of Neurology, Korea University Ansan Hospital|Departments of Neurology and Medical Science, Ewha Womans University School of Medicine and Ewha Medical Research Institute
Ansan, Republic of Korea|Seoul, Korea, Republic of
Department of Neurology, Korea University Ansan Hospital|Departments of Neurology and Medical Science, Ewha Womans University School of Medicine and Ewha Medical Research Institute
Ansan, Republic of Korea|Seoul, Korea, Republic of
Co-Author(s):
Younghan Lee, BEng
Department of Electrical and Computer Engineering (ECE) and Inter-University Semiconductor Research Center (ISRC), Seoul National University
Seoul, Korea, Republic of
Department of Electrical and Computer Engineering (ECE) and Inter-University Semiconductor Research Center (ISRC), Seoul National University
Seoul, Korea, Republic of
Youngseo Kim, MSc
Artificial Intelligence Convergence and Computational Medicine, System Health Science & Engineering, Graduate School of Ewha Womans University
Seoul, Korea, Republic of
Artificial Intelligence Convergence and Computational Medicine, System Health Science & Engineering, Graduate School of Ewha Womans University
Seoul, Korea, Republic of
Ho Bae, PhD
Artificial Intelligence Convergence and Computational Medicine, System Health Science & Engineering, Graduate School of Ewha Womans University
Seoul, Korea, Republic of
Artificial Intelligence Convergence and Computational Medicine, System Health Science & Engineering, Graduate School of Ewha Womans University
Seoul, Korea, Republic of
Yunheung Paek, PhD
Department of Electrical and Computer Engineering (ECE) and Inter-University Semiconductor Research Center (ISRC), Seoul National University
Seoul, Korea, Republic of
Department of Electrical and Computer Engineering (ECE) and Inter-University Semiconductor Research Center (ISRC), Seoul National University
Seoul, Korea, Republic of
Michael J. Prerau
Division of Sleep and Circadian Disorders, Brigham and Women’s Hospital
Boston, MA
Division of Sleep and Circadian Disorders, Brigham and Women’s Hospital
Boston, MA
Hyang Woon Lee, MD, PhD
Artificial Intelligence Convergence and Computational Medicine, System Health Science & Engineering, Graduate School of Ewha Womans University|Departments of Neurology and Medical Science, Ewha Womans University School of Medicine and Ewha Medical Research Institute
Seoul, Korea, Republic of|Seoul, Korea, Republic of
Artificial Intelligence Convergence and Computational Medicine, System Health Science & Engineering, Graduate School of Ewha Womans University|Departments of Neurology and Medical Science, Ewha Womans University School of Medicine and Ewha Medical Research Institute
Seoul, Korea, Republic of|Seoul, Korea, Republic of
Introduction:
Cognitive processing during sleep has been revealed to rely on the precisely synchronized interactions of rhythmic oscillation of neural activities. Current practice, however, significantly limits our ability to explore the continuously evolving neurophysiological dynamics by the low-resolution framework, which discretizes the neural states during sleep only into 5 categories with a temporal constraint of a non-overlapping 30-sec epoch. Recent studies have shown that the slow oscillation (SO) power can be used as a continuous correlate of depth of sleep. In this study, we aim to represent the personal intrinsic neural dynamic patterns during sleep by high-resolution phenotype mapping, which is independent of the temporal evolution of sleep architecture or night-to-night variability.
Methods:
We analyzed 80 polysomnography (PSG) data previously recorded from 20 middle-aged female subjects (mean age of 56.45 ± 3.97 years) for unrelated research, approved by the IRB of Ewha Womans University Mokdong Hospital and was registered with the Clinical Research Information Service (CRIS; study number KCT0001867). For all subjects, four PSGs at one-month intervals were recorded at the hospital sleep laboratory. We analyzed each night's EEG signal from the C3-A1 electrode derivation, and the data segment was from the time of lights out to the time of lights on.
SO-power was computed as the total power in the artifact-free EEG multitaper spectrogram between .5 – 2.0 Hz within a 60s window with 5s step (TW=30, L=29). Then, mean spectral power from .5-30 Hz was computed as a function of SO-power to obtain the state-based histogram. For dimensionality reduction, a low-resolution (60 x 41) representation of the SO-power histogram was created for each PSG recording.
To evaluate the intra-subject stability of the suggested framework, the similarity of the four SO-power histogram matrices from each participant was determined through mean square error (MSE) and Euclidian distance, which measures the average squared differences.
SO-power was computed as the total power in the artifact-free EEG multitaper spectrogram between .5 – 2.0 Hz within a 60s window with 5s step (TW=30, L=29). Then, mean spectral power from .5-30 Hz was computed as a function of SO-power to obtain the state-based histogram. For dimensionality reduction, a low-resolution (60 x 41) representation of the SO-power histogram was created for each PSG recording.
To evaluate the intra-subject stability of the suggested framework, the similarity of the four SO-power histogram matrices from each participant was determined through mean square error (MSE) and Euclidian distance, which measures the average squared differences.
Results:
The average length (time-in-bed) of our PSG recordings was 309.34 ± 46.41 minutes, and the sleep efficiency was 78.73 ± 10.85%. According to the standard guidelines of the AASM, 84.19 ± 6.94% was NREM sleep, of which the slow wave sleep was 4.85 ± 4.92%.
Despite night-to-night variability in sleep architecture evolution over time within each subject, the repeatedly calculated SO-power histograms from one's own four PSG recordings exhibited marked similar patterns, showing consistency in the neural dynamics of each individual. Moreover, we also confirmed the heterogeneity of the SO-power histogram between subjects. The MSE between the first night PSG recording and the other three PSG recordings obtained from the same subject was lower than the MSE between different subjects.
Despite night-to-night variability in sleep architecture evolution over time within each subject, the repeatedly calculated SO-power histograms from one's own four PSG recordings exhibited marked similar patterns, showing consistency in the neural dynamics of each individual. Moreover, we also confirmed the heterogeneity of the SO-power histogram between subjects. The MSE between the first night PSG recording and the other three PSG recordings obtained from the same subject was lower than the MSE between different subjects.
·Intra-individual stability and Inter-individual variability of the suggested framework
Conclusions:
In this study, we build on previous work (Kim et. al, Stokes et. al) and compute a SO-power histogram, which visualizes the median spectral power as a function of frequency and SO-power. In this work, we extend previous findings of night-to-night stability to show clear month-to-month stability. This allows us to observe how non-stationary oscillatory dynamics of the brain during sleep can comprehensively and quantitatively be represented using continuous-valued correlates of sleep depth without being affected by different temporal sleep evolution patterns.
By choosing a high-dimensional, low-variability representation of sleep EEG data, we can create a high-resolution phenotype mapping framework that is more sensitive to changes in nneural dynamic states, and thus a powerful tool for assessing potential changes in neurological health.
By choosing a high-dimensional, low-variability representation of sleep EEG data, we can create a high-resolution phenotype mapping framework that is more sensitive to changes in nneural dynamic states, and thus a powerful tool for assessing potential changes in neurological health.
Modeling and Analysis Methods:
EEG/MEG Modeling and Analysis 2
Perception, Attention and Motor Behavior:
Sleep and Wakefulness 1
Keywords:
Computational Neuroscience
Design and Analysis
Electroencephaolography (EEG)
NORMAL HUMAN
Sleep
1|2Indicates the priority used for review
Provide references using author date format
Prerau, M. J., Brown, R. E., Bianchi, M. T., Ellenbogen, J. M., & Purdon, P. L. (2017). Sleep neurophysiological dynamics through the lens of multitaper spectral analysis. Physiology, 32(1), 60-92.
Stokes, P. A., Rath, P., Possidente, T., He, M., Purcell, S., Manoach, D. S., ... & Prerau, M. J. (2023). Transient oscillation dynamics during sleep provide a robust basis for electroencephalographic phenotyping and biomarker identification. Sleep, 46(1), zsac223.
Kim, H. J., Kim, J., Lee, S., Kim, B., Kwon, E., Lee, J. E., ... & Lee, H. W. (2019). A double-blind, randomized, placebo-controlled crossover clinical study of the effects of alpha-s1 casein hydrolysate on sleep disturbance. Nutrients, 11(7), 1466.
Acknowledgments
This study was supported by grants from the Korea University Ansan Hospital [K2316061 to H.J.Kim], the Korea Health Technology R & D Project through the KHIDI, funded by the Ministry of Health & Welfare [HI19C1065 to H.J.Kim], the National Research Foundation of Korea [NRF-2019M3C1B8090803 and 2020R1A2C2013216 to H.W.Lee], the IITP grant through the MSIT/Ministry of Education & Future Planning [RS-2022-00155966 to H.W.Lee and RS-2023-00277326 to Y.Lee], and by the BK21 FOUR program of the Education and Research Program for Future ICT Pioneers programs of Ewha Womans University and Seoul National University in 2023.