Conference proceeding
Characterization of Arousals in Polysomnography Using the Statistical Significance of Power Change
2018 IEEE Signal Processing in Medicine and Biology Symposium (SPMB), pp.1-6
12/2018
DOI: 10.1109/SPMB.2018.8615603
Abstract
Arousals are neural events during sleep represented as abrupt increases of high-frequency electroencephalogram (EEG) signals in polysomnography (PSG). In clinical practice, a human scorer uses visual detection to demarcate the starts and ends of arousals, whereas other properties of arousals are hardly ever studied. Here we characterized arousals by the statistical significance of arousal-associated changes in the EEG signal power. To evaluate the test-retest reliability, we used a database of 1026 men who completed two PSGs separated by several years. Ten-second segments of EEG signals that either contained or were without arousals were extracted. For each segment, the power of EEG signal filtered in delta, theta, alpha, beta or gamma frequency band was computed. Then for each PSG, statistical significance of the difference in power between the arousal-containing and the arousal-absent group of EEG segments was computed. The statistical significance showed good test-retest reliability (intraclass correlation coefficient ICC0.40). In comparison, the numeric value of the difference in power showed generally poor test-retest reliability (ICC <0.10). The statistical significance had higher test-retest reliability in theta band (4-8 Hz) than in other frequency bands, and higher reliability in Stage 2 sleep than in rapid eye movement (REM) sleep. Furthermore, the statistical significance in theta band was not influenced by the incidence rate of arousals. Thus, statistical significance of power change in the theta band is a robust metric of arousal-associated EEG signal changes, which may become useful in studying diseases associated with abnormal arousals.
Details
- Title: Subtitle
- Characterization of Arousals in Polysomnography Using the Statistical Significance of Power Change
- Creators
- Junjie V Liu - Yale UniversityH. Klar Yaggi - Yale University
- Resource Type
- Conference proceeding
- Publication Details
- 2018 IEEE Signal Processing in Medicine and Biology Symposium (SPMB), pp.1-6
- Publisher
- IEEE
- DOI
- 10.1109/SPMB.2018.8615603
- ISSN
- 2372-7241
- eISSN
- 2473-716X
- Language
- English
- Date published
- 12/2018
- Academic Unit
- Neurology
- Record Identifier
- 9984303439902771
Metrics
4 Record Views