A Measure to Detect Sleep Onset Using Statistical Analysis of Spike Rhythmicity

A Measure to Detect Sleep Onset Using Statistical Analysis of Spike Rhythmicity

B.R. Purnima (Center for Medical Electronics and Computing, M.S. Ramaiah Institute of Technology, Bangalore, Karnataka, India), N. Sriraam (Center for Medical Electronics and Computing, M.S. Ramaiah Institute of Technology, Bangalore, Karnataka, India), U. Krishnaswamy (Department of Pulmonary Medicine, M.S. Ramaiah Hospitals, Bangalore, Karnataka, India) and K. Radhika (Department of Community Medicine, M.S. Ramaiah Medical College, Bangalore, Karnataka, India)
Copyright: © 2014 |Pages: 15
DOI: 10.4018/ijbce.2014010103
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

Electroencephalogram (EEG) signals derived from polysomnography recordings play an important role in assessing the physiological and behavioral changes during onset of sleep. This paper suggests a spike rhythmicity based feature for discriminating the wake and sleep state. The polysomnography recordings are segmented into 1 second EEG patterns to ensure stationarity of the signal and four windowing scheme overlaps (0%, 50%, 60% and 75%)of EEG pattern are introduced to study the influence of the pre-processing procedure. The application of spike rhythmicity feature helps to estimate the number of spikes from the given pattern with a threshold of 25%.Then non parametric statistical analysis using Wilcoxon signed rank test is introduced to evaluate the impact of statistical measures such as mean, standard deviation, p-value and box-plot analysis under various conditions .The statistical test shows significant difference between wake and sleep with p<0.005 for the applied feature, thus demonstrating the efficiency of simple thresholding in distinguishing sleep and wake stage .
Article Preview

Preprocessing

The polysomnography datasets from which Electroencephalogram is extracted is acquired from the sleep laboratory of M S Ramaiah Hospitals. As the EEG changes during onset of sleep occur first in the occipital lobe of the brain, channels O1 and O2 are considered for the data acquisition process. The sampling frequency of the acquired data is 256 Hz used for feature extraction. The system that is used for the analysis of the data is Sandman analysis system, the portion of which utilizes Primo Burner Technology © 2003-2006 Primo software. For the proposed study, three subjects’ recordings are considered for analysis. EEG being a non-stationary signal is segmented into 1 second patterns by applying different windowing as discussed below.

  • 1.

    Assuming that the given time series is divided into 1sec, there is no overlap between any two consecutive 1 second EEG patterns(SR1)

  • 2.

    Retaining the first segmented 1 second pattern, a 50% overlap is introduced in such a manner that the second EEG pattern comprise of 0.5 sec i.e. half of first EEG pattern and first half of second EEG pattern of SR1(SR2).

  • 3.

    The third and fourth windowing is based on 60% (SR3) and 75%(SR4) of the overlap. The procedure is same as discussed in (ii).

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 6: 2 Issues (2017)
Volume 5: 2 Issues (2016)
Volume 4: 2 Issues (2015)
Volume 3: 2 Issues (2014)
Volume 2: 2 Issues (2013)
Volume 1: 2 Issues (2012)
View Complete Journal Contents Listing