272 Long-term monitoring of trait-like characteristics of the sleep electroencephalogram using ear-EEG

• Published in: Sleep (New York, N.Y.) Vol. 44; no. Supplement_2; p. A109
• Main Authors: Hemmsen, Martin, Mikkelsen, Kaare, Rank, Mike, Kidmose, Preben
• Format: Journal Article
• Language: English
• Published: Westchester Oxford University Press 03.05.2021
• Subjects: Alzheimer's disease; Electroencephalography; NREM sleep; Signatures; Sleep
• ISSN: 0161-8105; 1550-9109
• DOI: 10.1093/sleep/zsab072.271

Introduction Wearable electroencephalogram (EEG) monitoring has a remarkable potential, it is safe, scalable and can track neural signatures for long periods. One such signature is the power spectra of non‐rapid-eye-movement (NREM) sleep which has been shown to demonstrate a trait‐like characteristic. Changes in personalized signatures has been associated with biomarkers of Alzheimer’s disease and is of great interest for early detection and clinical management. This work investigates monitoring of signatures using a wearable device that records EEG from the ear (ear-EEG) and compares the intra- and inter-individual similarity of the neural signatures with that from central scalp-EEG. Methods We initiated a two phased in-home study, monitoring 20 subjects for 4 nights (A), followed by a delayed but continued monitoring of 10 subjects for 12 nights (B). In A, subjects wore a dry-electrode ear-EEG system and a partial PSG, in B the subjects wore only the ear-EEG system. Subjects were instructed to follow their usual time schedule and lifestyle. Sleep stages were scored manually according to AASM in A and automatically in B. The grand average power spectra of NREM2 sleep were computed and log-transformed prior to calculating the cosine similarity for determination of the intra- and inter-individual similarity. Results The ear-EEG and scalp-EEG analysis showed that mean intra-individual similarity was higher than mean inter-individual similarity. Permutation tests indicate that the observed mean difference is statistically significant p<0.01 for both montages. Comparing the distributions of intra-individual similarities for ear-EEG and scalp-EEG, the observed mean difference is statistically significant p<0.05, in favor of a more stable ear-EEG signature. Comparing ear-EEG signatures between A and B, considering nights from A as reference, all subjects from B were most similar with its own reference signature. Considering signatures from individual nights the accuracy paring subjects from A and B were 88% correct. Conclusion Nocturnal ear-EEG measures trait-like characteristics as reliable as scalp-EEG. The neural signature is stable over time within healthy subjects and demonstrated its ability as a personalized signature. Support (if any):