Sridhar Vijendra

Time domain analysis of R-wave attenuation envelope for sleep apnea detection

Time domain analysis was carried out on the R-wave attenuation (RWA) envelope of the subjects with and without obstructive sleep apnea. The RWA envelope is derived from the morphology of the electrocardiogram (ECG) obtained during polysomnography data collection of the subjects. Nocturnal polysomnography was performed on 16 normal subjects and 14 obstructive sleep apnea (OSA) patients. The ECG from the polysomnography data was divided into fifteen minute epochs for analysis. The QRS detection was carried out by an algorithm using Hilbert transform. Standard deviation of each of the fifteen one minute epochs in fifteen minute epoch of the RWA envelope was calculated. Standard deviation of these fifteen parameters was observed to have considerably good sensitivity and specificity to sleep apnea. For the clips selected from normal subjects, the parameter produced a sensitivity of 78.57% and specificity of 70.33% for the training set and sensitivity of 87.5% and specificity of 80.95 for the testing set. For the clips selected from OSA subjects, the parameter produced a sensitivity of 72.46% and specificity of 73.53% for the training set and sensitivity of 82.86% and specificity of 66.67% for the testing set.

The use of R-wave morphology in the detection of sleep-disordered breathing using the electrocardiogram - a comparison between leads

Power spectral analysis of time series derived from the R-wave morphology of the ECG was employed to identify a suitable lead configuration for the detection of sleep-disordered breathing (SDB) using the electrocardiogram (ECG). 16 subjects (46 /spl plusmn/ 9.2 yrs, 8 males), who did not report problems during sleep, and 13 subjects previously diagnosed with SDB (49 /spl plusmn/ 8.8 yrs, 7 males) underwent an overnight sleep study at an accredited sleep center. Power values derived from the spectra of the R-peaks envelope were tested for their sensitivity and specificity in discriminating between epochs containing normal breathing from epochs containing SDB. Of the three tested lead configurations using two parameters NB1 and NB2 derived from the power spectrum, lead I produced the best results with a sensitivity of 92.8% and a specificity of 88.0% for the case of parameter NB1 and a sensitivity of 85.7% and a specificity of 76.0% for the case of parameter NB2.

Frequency domain analysis of heart rate variability in sleep disordered breathing

Frequency domain analysis of the R-R intervals was performed in subjects with and without sleep disordered breathing (SDB). Data from Physionets Apnea-ECG database as well as data collected at our sleep disorders clinic were used for this study. 15-minute ECG epochs were selected from polysomnography data. Power spectrum analysis of the resampled R-R interval series (1.2 Hz) was performed using Welchs averaged periodogram method. Normalized power in two bands, AB/sub 1/ (0.019 to 0.071 Hz) and AB/sub 2/ (0.019 to 0.036 Hz) were used to compare the impact of SDB events on the spectrum of R-R intervals. Normalized AB/sub 1/ power produced a sensitivity of 76% and a specificity of 100% in the data from the Apnea-ECG database. The same parameter produced a sensitivity of 76% and specificity of 52% in the data collected at our sleep laboratory. This difference in the results may be attributed to the severity and duration of SDB events present in the two datasets.

Computer-based analysis of heart rate variability signal for detection of sleep disordered breathing in children

A computer-based analysis system was developed to display and analyze heart rate variability (HRV). ECG, oxygen saturation and respiratory signals (airflow, abdominal and thoracic movements), were used as raw data. The heart rate variability signal was derived from ECG by applying a Hilbert transform-based algorithm for reliable QRS complex detection. Following the guidelines suggested by the Task Force of the European Society of Cardiology and the North American Society for Pacing and Electrophysiology, appropriate time-domain and frequency-domain methods were used for HRV signal analysis. Autoregressive modeling of the HRV power spectrum was achieved by implementing the Burg algorithm. Three main spectral features were clearly distinguished in the heart rate variability signal spectrum from polysomnographic recordings of different sleep stages and were correlated with respiratory parameters. The integrated graphical user interface was developed using LabView and the signal processing algorithms were implemented using Matlab application programs. In this paper we present an overview of the system and analyze pilot data for two children undergoing nocturnal polysomnography. The pilot data demonstrated that the HRV analysis system may potentially distinguish between periods of normal and sleep disordered breathing (SDB) in children.

Classification of Cheyne-Stokes breathing and obstructive sleep apnea using ECG.

High cost of diagnostic studies to detect sleep disordered breathing and lack of availability of certified sleep laboratories in all inhabited areas make investigation of alternative methods of detecting sleep disordered breathing attractive. This study aimed to explore the possibility of discerning obstructive sleep apnea (OSA) from Cheyne-Stokes respiration (CSR) using overnight electrocardiography (ECG). Polysomnographic and ECG signals were acquired from the 13 OSA and 7 CSR volunteer subjects. Two signals: R-Wave Attenuation (RWA) and Heart Rate Variability (HRV) series were derived from the ECG. Using frequency domain analysis, various frequency bands in the power spectrum of RWA and HRV signals were identified that showed sensitivity to OSA and CSR events. A three-stage algorithm was developed to detect and differentiate OSA events from CSR events using RWA and HRV analysis. To test the algorithm, the ECG data was divided into fifteen minute epochs for analysis. Seventy two epochs containing OSA and 72 with CSR events were selected. 48 OSA clips and 48 CSR clips were randomly selected to form the training set. The remaining 24 clips in each category formed the test set. This method produced an average sensitivity of 95.83% and specificity of 79.16% in the training set and sensitivity of 87.5% and a specificity of 75% in the test set

Spectral Analysis of R Wave Attenuation and Heart Rate Variability for Detection of Cheyne Stokes Breathing

Spectral analysis was carried out on the R-wave attenuation (RWA) trend and heart rate variability (HRV) series, derived from the polysomnographic electrocardiogram (ECG) of the subjects with and without Cheyne Stokes breathing. Nocturnal polysomnography was performed on 16 normal subjects and 7 subjects with Cheyne Stokes breathing (CSB) patients. The polysomnographic ECG data was divided into fifteen minute epochs for analysis. These epochs are processed to obtain the RWA. Hilbert transform based algorithm was used for QRS detection. Power spectrum of RWA and HRV are computed for each clip by using Welchs averaged periodogram method. HRV is sensitive to REM sleep as well and hence not specific to sleep apnea. Hence the parameters derived from HRV alone cannot be used as diagnostic markers. Hence a combined detection scheme which uses parameters derived from RWA and HRV power spectrum is used in the proposed method to increase detection accuracy. This method produced a sensitivity of 84.75% and specificity of 87.03% in the training set and sensitivity of 85.78% and a specificity of 87.19% in the test set

ECG Biomarkers for Simultaneous Detection of Obstructive Sleep Apnea and Cheyne-Stokes Breathing

Use of extended electrocardiography (ECG) for detection of sleep disordered breathing SDB when obstructive sleep apneas and Cheyne-Stokes breathing are simultaneously present is explored. A multi-tier algorithm is designed that uses quantitative changes in the morphology of the QRS complex of Lead 1 and V4 due of SDB events and combines those changes with variations in heart rate to detect each type of SDB. For this purpose, ECG signals are divided into 15 minute epochs. These epochs are then subjected to baseline wander removal and R peak detection. An envelope of R peaks is computed to derive R wave attenuation (RWA). Concurrently, the heart rate variability (HRV) is also computed. Various biomarkers derived from these trends are combined to develop an algorithm to classify Normal, OSA and CSR epochs. One hundred and five (105) data clips from 15 subjects were used to test the proposed algorithm. It produced detection rates of 93.75%, 100% and 83.3% for Normal, OSA and CSR epochs respectively in case of training set (66 clips). Detection rates of 75%, 85.71% and 70.5% for Normal, OSA and CSR epochs respectively were obtained in case of test set (39 clips).