Systematic comparison of different algorithms for apnoea detection based on electrocardiogram recordings

Sleep apnoea is a common disorder that is usually diagnosed through expensive studies conducted in sleep laboratories. Sleep apnoea is accompanied by a characteristic cyclic variation in heart rate or other changes in the waveform of the electrocardiogram (ECG). If sleep apnoea could be diagnosed using only the ECG, it could be possible to diagnose sleep apnoea automatically and inexpensively from ECG recordings acquired in the patient's home. This study had two parts. The first was to assess the ability of an overnight ECG recording to distinguish between patients with and without apnoea. The second was to assess whether the ECG could detect apnoea during each minute of the recording. An expert, who used additional physiological signals, assessed each of the recordings for apnoea. Research groups were invited to access data via the world-wide web and submit algorithm results to an international challenge linked to a conference. A training set of 35 recordings was made available for algorithm development, and results from a test set of 35 different recordings were made available for independent scoring. Thirteen algorithms were compared. The best algorithms made use of frequency-domain features to estimate changes in heart rate and the effect of respiration on the ECG waveform. Four of these algorithms achieved perfect scores of 100% in the first part of the study, and two achieved an accuracy of over 90% in the second part of the study.     

Unconstrained detection of respiration rhythm and pulse rate with one under-pillow sensor during sleep

A completely non-invasive and unconstrained method is proposed to detect respiration rhythm and pulse rate during sleep. By employing wavelet transformation (WT), waveforms corresponding to the respiration rhythm and pulse rate can be extracted from a pulsatile pressure signal acquired by a pressure sensor under a pillow. The respiration rhythm was obtained by an upward zero-crossing point detection algorithm from the respiration-related waveform reconstructed from the WT 2⁶ scale approximation, and the pulse rate was estimated by a peak point detection algorithm from the pulse-related waveform reconstructed from the WT 2⁴ and 2⁵ scale details. The finger photo-electric plethysmogram (FPP) and nasal thermistor signals were recorded simultaneously as reference signals. The reference pulse rate and respiration rhythm were detected with the peak and upward zero-crossing point detection algorithm. This method was verified using about 24 h of data collected from 13 healthy subjects. The results showed that, compared with the reference data, the average error rates were 3.03% false negative and 1.47% false positive for pulse rate detection in the extracted pulse waveform. Similarly, 4.58% false negative and 3.07% false positive were obtained for respiration rhythm detection in the extracted respiration waveform. This study suggests that the proposed method is suitable, in sleep monitoring, for the diagnosis of sleep apnoea or sudden death syndrome.     

由单通道心电提取呼吸信息的算法

基于除呼吸作用以外，能够引起正常的QRS波群中R波振幅的干扰因素也较少的事实，提出使用室上性的QRS波群由心电提取呼吸信息（EDR）（ECG-derived respiratory signal）的算法。该算法先将单通道心电进行小波分解，再将小波近似在适当尺度上的室上性波R波振幅在时间上进行延拖，通过一低通滤波器后再将抽样频率降低至5Hz，再通过一带通滤波器（0．1～0．4Hz）后，即可得到EDR。     

QRS波群时频检测方法的新进展

在心电信号的QRS波群检测算法中,最为有效的是把待测信号在时域或频域内进行适当的变换,以分离或加强QRS分量,抑制各种噪声干扰,再进行判决定位。简要回顾了以滤波器为代表的单一尺度时频变换方法、以小波分析为代表的多尺度时频变换方法,以及另一种较新的多尺度时频变换分析方法—经验模式分解(EMD)等在QRS波群检测中的应用情况和发展前景。     

Automated recognition of patients with obstructive sleep apnoea using wavelet-based features of electrocardiogram recordings

Patients with obstructive sleep apnoea syndrome (OSAS) are at increased risk of developing hypertension and other cardiovascular diseases. This paper explores the use of support vector machines (SVMs) for automated recognition of patients with OSAS types (±) using features extracted from nocturnal ECG recordings, and compares its performance with other classifiers. Features extracted from wavelet decomposition of heart rate variability (HRV) and ECG-derived respiration (EDR) signals of whole records (30 learning sets from physionet) are presented as inputs to train the SVM classifier to recognize OSAS± subjects. The optimal SVM parameter set is then determined by using a leave-one-out procedure. Independent test results have shown that an SVM using a subset of a selected combination of HRV and EDR features correctly recognized 30/30 of physionet test sets. In comparison, classification performance of K-nearest neighbour, probabilistic neural network, and linear discriminant classifiers on test data was lower. These results, therefore, demonstrate considerable potential in applying SVM in ECG-based screening and can aid sleep specialists in the initial assessment of patients with suspected OSAS.     

The effect of time of day on apnoea index in the sleeping rat

This study tested the hypothesis that apnoea index would be greater during daytime sleep than nighttime sleep in the rat. Electroencephalogram and electromyogram were monitored via biotelemetry implant and respiration was measured using whole body plethysmography in six male rats in two separate 34h recording sessions per animal. Apnoeas were classified as "spontaneous" or "post-sigh". Daily average spontaneous apnoea index was 35 times greater (p<0.0001) during rapid eye movement (REM) sleep than in non-REM (NREM) sleep. In contrast, daily average post-sigh apnoea index was not significantly greater in REM sleep than in non-REM (NREM) sleep (p=0.39). There was a greater post-sigh apnoea index during daytime REM than during nighttime REM (p=0.043) but REM-related spontaneous apnoea index was unaffected by time of day. There was no day to night difference in spontaneous apnoea index or post-sigh apnoea index during NREM sleep. Respiratory variability (coefficient of variation for breath duration and tidal volume) was not affected by time of day in REM or NREM sleep. We conclude that the circadian timing system has no effect on apnoea index during NREM sleep in the rat, but it may influence the propensity for post-sigh apnoea during REM sleep.     

A behavioural test to assess daytime sleepiness in obstructive sleep apnoea

Daytime sleepiness is an important symptom of obstructive sleep apnoea syndrome (OSAS). The standard tests for its objective quantification use EEG recordings, and are time consuming and expensive, which makes them difficult to use for large studies. This study assesses the ability of a simple test of sustained ‘wakefulness’ to discriminate the excessive somnolence of severe symptomatic obstructive sleep apnoea from normality, and compares its results to the traditional EEG based Maintenance of Wakefulness Test (MWT). Ten subjects (7M 3F) with severe sleep apnoea (>4% SaO₂ dip rate median 32.7 (90% central range 9.7–65.6)) and symptoms of daytime sleepiness, (Epworth Sleepiness Score (ESS)17(10–24)) and 10 normal subjects (4M 6F, ESS 3.5(1–8)) were studied. The MWT and the behavioural test (Oxford SLEep Resistance test – OSLER test) were performed on each subject in random order on 2 separate days. The protocol for both tests was the same with 4 × 40 min sleep resistance challenges throughout the day while sound isolated in a darkened room. During the OSLER test subjects were asked to press a switch in response to a light emitting diode (LED), which was lit for 1 s in every three. Both the switch and the light were connected to a computer that stored both the number of times the light was illuminated and whether a correct response was made. The OSLER test discriminated the normal subjects from the sleep apnoea group (mean sleep latency (min) normal group 39.8, OSA group 10.5) as well as the traditional MWT (normal group 38.1, OSA group 7.3) and was much simpler to administer. This test has the advantage that sleep onset is defined objectively and automatically as a failure to respond to the light, rather than from EEG interpretation, which is inevitably partly subjective. This technique may provide a simple and robust method of objectively quantifying daytime sleepiness for large studies.     

Assessment of sleep‐disordered breathing using a non‐contact bio‐motion sensor

Obstructive sleep apnoea is a highly prevalent but under‐diagnosed disorder. The gold standard for diagnosis of obstructive sleep apnoea is inpatient polysomnography. This is resource intensive and inconvenient for the patient, and the development of ambulatory diagnostic modalities has been identified as a key research priority. SleepMinder (BiancaMed, NovaUCD, Ireland) is a novel, non‐contact, bedside sensor, which uses radio‐waves to measure respiration and movement. Previous studies have shown it to be effective in measuring sleep and respiration. We sought to assess its utility in the diagnosis of obstructive sleep apnoea. SleepMinder and polysomnographic assessment of sleep‐disordered breathing were performed simultaneously on consecutive subjects recruited prospectively from our sleep clinic. We assessed the diagnostic accuracy of SleepMinder in identifying obstructive sleep apnoea, and how SleepMinder assessment of the apnoea–hypopnoea index correlated with polysomnography. Seventy‐four subjects were recruited. The apnoea–hypopnoea index as measured by SleepMinder correlated strongly with polysomnographic measurement (r = 0.90; P ≤ 0.0001). When a diagnostic threshold of moderate–severe (apnoea–hypopnoea index ≥15 events h^?1) obstructive sleep apnoea was used, SleepMinder displayed a sensitivity of 90%, a specificity of 92% and an accuracy of 91% in the diagnosis of sleep‐disordered breathing. The area under the curve for the receiver operator characteristic was 0.97. SleepMinder correctly classified obstructive sleep apnoea severity in the majority of cases, with only one case different from equivalent polysomnography by more than one diagnostic class. We conclude that in an unselected clinical population undergoing investigation for suspected obstructive sleep apnoea, SleepMinder measurement of sleep‐disordered breathing correlates significantly with polysomnography.     

Energy based feature extraction for classification of sleep apnea syndrome

In this paper it is aimed to classify sleep apnea syndrome (SAS) by using discrete wavelet transforms (DWT) and an artificial neural network (ANN). The abdominal and thoracic respiration signals are separated into spectral components by using multi-resolution DWT. Then the energy of these spectral components are applied to the inputs of the ANN. The neural network was configured to give three outputs to classify the SAS situation of the subject.The apnea can be mainly classified into three types: obstructive sleep apnea (OSA), central sleep apnea (CSA) and mixed sleep apnea (MSA). During OSA, the airway is blocked while respiratory efforts continue. During CSA the airway is open, however, there are no respiratory efforts. In this paper we aim to classify sleep apnea in one of three basic types: obstructive, central and mixed. A significant result was obtained.     

Suppression of electromyogram interference on the electrocardiogram by transform domain denoising

A method for suppression of electromyogram (EMG) interference in electrocardiogram (ECG) recordings is presented. By assuming that the EMG is long-term non-stationary Gaussian noise, two successive decompositions were proposed, and the data transformed for Wiener filtering. Successive ECG cycles were rearranged and aligned by the R-wave, forming a matrix containing separated heart cycles in its rows. A short-window discrete cosine transform (DCT) was applied to the columns of the matrix for inter-cycle de-correlation. Next, Weiner filtering in a translation-invariant wavelet domain was performed on the DCT-transformed matrix rows for de-correlation of the data into each ECG cycle. The method resulted in an improvement in the signal-to-noise ratio of more than 10 db, a threefold reduction in mean relative amplitude errors and reduced ripple artifacts around the signal transients, thus preserving the waveform in diagnostically important signal segments.     

ECG信号小波变换与峰谷检测算法的研究

本文在ECG信号检测过程中，将ECG信号在3尺度上的Haar小波分解的细节信号模极大值对检测与数学形态学峰谷检测相结合，提出了ECG信号小波变换与峰谷检测算法，该算法弥补了小波变换算法对ECG信号时域特征检测的不足，有效地提高了ECG信号检测的准确度。     

A new hybrid algorithm for ECG signal compression based on the wavelet transformation of the linearly predicted error.

This paper describes a hybrid technique based on the combination of wavelet transform and linear prediction to achieve very effective electrocardiogram (ECG) data compression. First, the ECG signal is wavelet transformed using four different discrete wavelet transforms (Daubechies, Coiflet, Biorthogonal and Symmlet). All the wavelet transforms are based on dyadic scales and decompose the ECG signals into five detailed levels and one approximation. Then, the wavelet coefficients are linearly predicted, where the error corresponding to the difference between these coefficients and the predicted ones is minimized in order to get the best predictor. In particular, the residuals of the wavelet coefficients are uncorrelated and hence can be represented with fewer bits compared to the original signal. To further increase the compression rate, the residual sequence obtained after linear prediction is coded using a newly developed coding technique. As a result, a compression ratio (Cr) of 20 to 1 is achieved with percentage root-mean square difference (PRD) less than 4%. The algorithm is compared to an alternative compression algorithm based on the direct use of wavelet transforms. Experiments on selected records from the MIT-BIH arrhythmia database reveal that the proposed method is significantly more efficient in compression. The proposed compression scheme may find applications in digital Holter recording, in ECG signal archiving and in ECG data transmission through communication channels.     

Research of fetal ECG extraction using wavelet analysis and adaptive filtering

Extracting clean fetal electrocardiogram (ECG) signals is very important in fetal monitoring. In this paper, we proposed a new method for fetal ECG extraction based on wavelet analysis, the least mean square (LMS) adaptive filtering algorithm, and the spatially selective noise filtration (SSNF) algorithm. First, abdominal signals and thoracic signals were processed by stationary wavelet transform (SWT), and the wavelet coefficients at each scale were obtained. For each scale, the detail coefficients were processed by the LMS algorithm. The coefficient of the abdominal signal was taken as the original input of the LMS adaptive filtering system, and the coefficient of the thoracic signal as the reference input. Then, correlations of the processed wavelet coefficients were computed. The threshold was set and noise components were removed with the SSNF algorithm. Finally, the processed wavelet coefficients were reconstructed by inverse SWT to obtain fetal ECG. Twenty cases of simulated data and 12 cases of clinical data were used. Experimental results showed that the proposed method outperforms the LMS algorithm: (1) it shows improvement in case of superposition R-peaks of fetal ECG and maternal ECG; (2) noise disturbance is eliminated by incorporating the SSNF algorithm and the extracted waveform is more stable; and (3) the performance is proven quantitatively by SNR calculation. The results indicated that the proposed algorithm can be used for extracting fetal ECG from abdominal signals.     

Cross-correlation of EEG frequency bands and heart rate variability for sleep apnoea classification

Sleep apnoea is a sleep breathing disorder which causes changes in cardiac and neuronal activity and discontinuities in sleep pattern when observed via electrocardiogram (ECG) and electroencephalogram (EEG). Using both statistical analysis and Gaussian discriminative modelling approaches, this paper presents a pilot study of assessing the cross-correlation between EEG frequency bands and heart rate variability (HRV) in normal and sleep apnoea clinical patients. For the study we used EEG (delta, theta, alpha, sigma and beta) and HRV (LF_nu, HF_nu and LF/HF) features from the spectral analysis. The statistical analysis in different sleep stages highlighted that in sleep apnoea patients, the EEG delta, sigma and beta bands exhibited a strong correlation with HRV features. Then the correlation between EEG frequency bands and HRV features were examined for sleep apnoea classification using univariate and multivariate Gaussian models (UGs and MGs). The MG outperformed the UG in the classification. When EEG and HRV features were combined and modelled with MG, we achieved 64% correct classification accuracy, which is 2 or 8% improvement with respect to using only EEG or ECG features. When delta and acceleration coefficients of the EEG features were incorporated, then the overall accuracy improved to 71%.     

基于小波变换与形态学运算的ECG综合检测算法的研究

针对心电波形检测中小波变换算法的缺点 ,在 ECG特征点检测中 ,将原始信号在 3尺度上的 haar小波分解的细节信号模极大值对检测法与数学形态学峰谷检测相结合 ,提出了一种新的心电波形特征点综合检测算法 ,该算法弥补了小波变换算法对信号振幅检测上的不足 ,有效地提高了心电信号特征点检测的准确度。     

An improved algorithm for respiration signal extraction from electrocardiogram measured by conductive textile electrodes using instantaneous frequency estimation

In this paper, an improved algorithm for the extraction of respiration signal from the electrocardiogram (ECG) in home healthcare is proposed. The whole system consists of two-lead electrocardiogram acquisition using conductive textile electrodes located in bed, baseline fluctuation elimination, R-wave detection, adjustment of sudden change in R-wave area using moving average, and optimal lead selection. In order to solve the problems of previous algorithms for the ECG-derived respiration (EDR) signal acquisition, we are proposing a method for the optimal lead selection. An optimal EDR signal among the three EDR signals derived from each lead (and arctangent of their ratio) is selected by estimating the instantaneous frequency using the Hilbert transform, and then choosing the signal with minimum variation of the instantaneous frequency. The proposed algorithm was tested on 15 male subjects, and we obtained satisfactory respiration signals that showed high correlation (r ² > 0.8) with the signal acquired from the chest-belt respiration sensor.     

基于功率谱特征的心电数据小波压缩方法

利用心电功率谱特征,探索心电数据压缩新方法。用小波分解心电信号为高频与低频分量,对低频分量继续分解达到要求的级数,对高频分量则根据其所在频段的能量,对临床诊断的价值加以取舍。对MIT生理信号数据库心电数据的压缩与还原分析表明,该方法平衡了压缩比与还原精度之间的矛盾,既具有较高的压缩比,又具有较高的还原精度,而且对信号的适应性也明显增强。另外,该压缩方法还具有一定的去噪作用。说明结合心电功率谱特征与小波变换方法压缩心电有其优势。     

Differences in relationships among sleep apnoea, glucose level, sleep duration and sleepiness between persons with and without type 2 diabetes

Obstructive sleep apnoea is common in patients with diabetes. Recently, it was reported that short sleep duration and sleepiness had deleterious effects on glucose metabolism. Thereafter, several reports showed relationships between glucose metabolism and obstructive sleep apnoea, sleep duration or sleepiness. But the interrelationships among those factors based on recent epidemiological data have not been examined. We analysed data on 275 male employees (age, 44±8years; body mass index, 23.9±3.1kg m(-2) ) who underwent a cross-sectional health examination in Japan. We measured fasting plasma glucose, sleep duration using a sleep diary and an actigraph for 7days, and respiratory disturbance index with a type 3 portable monitor for two nights. Fifty-four subjects (19.6%) had impaired glucose metabolism, with 21 having diabetes. Of those 21 (body mass index, 25.9±3.8kgm(-2) ), 17 (81.0%) had obstructive sleep apnoea (respiratory disturbance index≥5). Regarding the severity of obstructive sleep apnoea, 10, four and three had mild, moderate and severe obstructive sleep apnoea, respectively. The prevalence of obstructive sleep apnoea was greater in those with than without diabetes (P=0.037). Multiple regression analyses showed that the respiratory disturbance index independently related to fasting plasma glucose only in the diabetic subjects. In patients with diabetes, after adjustment for age, waist circumference, etc. sleep fragmentation had a greater correlation with fasting plasma glucose than sleep duration, but without significance (P=0.10). Because the prevalence of obstructive sleep apnoea is extremely high in patients with diabetes, sufficient sleep duration with treatment for obstructive sleep apnoea, which ameliorates sleep fragmentation, might improve fasting plasma glucose.     

Biomedical signal processing (in four parts)

This is the second in a series of four tutorial papers on biomedical signal processing, and it concerns the relationships between commonly used frequency transforms. It begins with the Fourier series and Fourier transform for continuous time signals and extends these concepts for aperiodic discrete time data and then periodic discrete time data. The Laplace transform is discussed as an extension of the Fourier transform. The z-transform is introduced and the ideas behind the chirp-z transform are described. The equivalence between the time and frequency domains is described in terms of Parseval's theorem and the theory of convolution. The use of the FFT for fast convolution and fast correlation is described for both short recordings and long recordings that must be processed in sections.     

Automatic analysis of electro-encephalogram sleep spindle frequency throughout the night

A fully automatic method to analyse electro-encephalogram (EEG) sleep spindle frequency evolution during the night was developed and tested. Twenty allnight recordings were studied from ten healthy control subjects and ten sleep apnoea patients. A total of 22 868 spindles were detected. The overall mean spindle frequency was significantly higher in the control subjects than in the apnoea patients (12.5Hz against 11.7Hz, respectively; p<0.004). The proposed method further identified the sleep depth cycles, and the mean spindle frequency was automatically determined inside each sleep depth cycle. In control subjects, the mean spindle frequency increased from 12.0Hz in the first sleep depth cycle to 12.6Hz in the fifth cycle. No such increase was observed in the sleep apnoea patients. This difference in the spindle frequency evolution was statistically significant (p<0.004). The advantage of the method is that no EEG amplitude thresholds are needed.