Removing electroencephalographic artifacts by blind source separation

Eye movements, eye blinks, cardiac signals, muscle noise, and line noise present serious problems for electroencephalographic (EEG) interpretation and analysis when rejecting contaminated EEG segments results in an unacceptable data loss. Many methods have been proposed to remove artifacts from EEG recordings, especially those arising from eye movements and blinks. Often regression in the time or frequency domain is performed on parallel EEG and electrooculographic (EOG) recordings to derive parameters characterizing the appearance and spread of EOG artifacts in the EEG channels. Because EEG and ocular activity mix bidirectionally, regressing out eye artifacts inevitably involves subtracting relevant EEG signals from each record as well. Regression methods become even more problematic when a good regressing channel is not available for each artifact source, as in the case of muscle artifacts. Use of principal component analysis (PCA) has been proposed to remove eye artifacts from multichannel EEG. However, PCA cannot completely separate eye artifacts from brain signals, especially when they have comparable amplitudes. Here, we propose a new and generally applicable method for removing a wide variety of artifacts from EEG records based on blind source separation by independent component analysis (ICA). Our results on EEG data collected from normal and autistic subjects show that ICA can effectively detect, separate, and remove contamination from a wide variety of artifactual sources in EEG records with results comparing favorably with those obtained using regression and PCA methods. ICA can also be used to analyze blink-related brain activity.     

Extracting sources from noisy abdominal phonograms: a single-channel blind source separation method

In this work we highlight a methodology that extracts sources from noisy single-channel abdominal phonograms. First, an appropriate matrix of delays is constructed. Next, multiple independent components are calculated using the FastICA algorithm. Then these components are projected back to the measurement space and classified for recovering the sources of interest. Single-channel phonograms obtained from three different subjects were analysed. Results show successful extraction of foetal heart sounds (FHS), maternal respiration/pulse wave, and line noise. It is important to point out the high performance of the method for extracting the former two as separate sources; especially due to the fact that pulse wave and FHS may overlap as maternal and foetal QRSs do in the abdominal ECG. The most outstanding factor is that this is achieved using a single-channel method. So, this approach extracts physiological sources from noisy abdominal phonograms, and we believe it will be useful for surveillance, not only for foetal well-being but also for maternal condition.     

Characterisation of acute myocardial ischaemia in a canine model based on principal component analysis of unipolar endocardial electrograms

The study presents a method for identifying endocardial electrical features relevant to local ischaemia detection at rest. The method consists of, first, normalisation of electrograms to a uniform representation; secondly, the use of principal component analysis to reduce the dimensionality of the electrogram vector space; and, thirdly, a search for a classification axis that matches the degree of ischaemia present in the tissue. Left ventricular myocardial states were assessed by echocardiography and NOGA mapping in eight dogs at baseline and then immediately after, 5h after and 3 days after occlusion of the left anterior descending coronary artery. Five principal components were required to approximate electrograms with an average error of less than 10% of the peak-to-peak amplitude. Correlations of 0.77, 0.80 and 0.84 were obtained between the principal component-based parameters and the echocardiography scores at the three ischaemic stages, respectively. Expression of these parameters in the time domain showed that the major changes occurred in the depolarisation segment of the endocardial electrogram as well as in the ST-segment. In conclusion, the proposed method provides a suitable alternative co-ordinate system for the classification of ischaemic regions and highlights signal segments that change as a result of pathology.     

心电图自动诊断系统的研制

本文利用数字信号处理和波形识别技术，对心电图的ＱＲＳ波、Ｐ波、Ｔ波的时段进行了计算机的自动检测并对室性期前收缩等１５种异常心电进行自动诊断。打印输出相应的诊断报告。为了检验本系统的稳定性和可信度，本文还利用美国麻省理工学院的ＭＩＴ－ＢＩＨ数据库对本文所使用的方法进行了检测，取得了较好的效果。     

Automatic removal of eye movement and blink artifacts from EEG data using blind component separation

Signals from eye movements and blinks can be orders of magnitude larger than brain-generated electrical potentials and are one of the main sources of artifacts in electroencephalographic (EEG) data. Rejecting contaminated trials causes substantial data loss, and restricting eye movements/blinks limits the experimental designs possible and may impact the cognitive processes under investigation. This article presents a method based on blind source separation (BSS) for automatic removal of electroocular artifacts from EEG data. BBS is a signal-processing methodology that includes independent component analysis (ICA). In contrast to previously explored ICA-based methods for artifact removal, this method is automated. Moreover, the BSS algorithm described herein can isolate correlated electroocular components with a high degree of accuracy. Although the focus is on eliminating ocular artifacts in EEG data, the approach can be extended to other sources of EEG contamination such as cardiac signals, environmental noise, and electrode drift, and adapted for use with magnetoencephalographic (MEG) data, a magnetic correlate of EEG.     

Blind separation of multichannel electrogastrograms using independent component analysis based on a neural network

The electrogastrogram (EGG) is an abdominal surface measurement of gastric myo-electrical activity which regulates gastric contractions. It is of great clinical importance to record and analyse multichannel EGGs, which provide more information on the propagation and co-ordination of gastric contractions. EGGs are, however, contaminated by myo-electric interference from other organs and artefacts such as motion and respiration. The aim of the study is to separate the gastric signal from noisy multichannel EGGs without any information on the interference, using independent component analysis. A neural-network model is proposed, and corresponding unsupervised learning algorithms are developed to achieve the separation. The performance of the proposed method is investigated using artificial data simulating real EGG signals. Experimental EGG data are obtained from humans and dogs. The processed results of both simulated and real EGG data show the following: first, the proposed method is able to separate normal gastric slow waves from respiratory artefacts and random noises. It is also able to extract gastric slow waves, even when the EGG is contaminated by severe respiratory and ECG artefacts. Secondly, when the stomach contains various gastric electric signals with different frequencies, the proposed method is able to separate these different signals, as illustrated by simulations. These data suggest that the proposed method can be used to separate gastric slow waves, respiratory and motion artefacts, and intestinal myo-electric interference that are mixed in the EGG. It can also be used to detect gastric slow-wave uncoupling, during which the stomach has multiple gastric signals with different frequencies. It is believed that the proposed method may also be applicable to other biomedical signals.     

Estimation of atrial fibrillatory wave from single-lead atrial fibrillation electrocardiograms using principal component analysis concepts

A new method for the assessment of the atrial fibrillatory wave (AFW) from the ECG is presented. This methodology is suitable for signals registered from Holter systems, where the reduced number of leads is insufficient to exploit the spatial information of the ECG. The temporal dependence of the bio-electrical activity were exploited using principal component analysis. The main features of ventricular and atrial activity were extracted, and several basis signals for each subspace were determined. Hence, the estimated (AFW) are reconstructed exclusively from the basis signals that formed the atrial subspace. Its main advantage with respect to adaptive template subtraction techniques was its robustness to variations in the QRST morphology, which thus minimised QRST residua. The proposed approach was first validated using a database of simulated recordings with known atrial activity content. The estimated AFW was compared with the original AFW, obtaining correlation indices of 0.774±0.106. The suitability of this methodology for real recordings was also proven, though its application to a set of paroxysmal AF ECGs. In all cases, it was possible to detect the main frequency peak, which was between 4.6 Hz and 6.9 Hz for the patients under study.     

Detection of previously unrecognized daytime desaturation in children with chronic lung disease

Primary objective: The prime rationale of this research is to investigate the possible occurrence of previously unrecognized episodes of desaturation apparent in preterm infants with chronic lung disease as they freely move around a non-artificial environment.

Research design: The study comprises 58 hours of telemetric recordings of SpO₂, heart rate, body movement and temperature, along with full ECG and photoplethysmographic waveforms for eight preterm subjects in their home environment.

Main outcome/results: The data is analysed for remarkable events, more particularly periods of spontaneous desaturation. Statistical results for all case studies are collated into a table along with examples of graphical analysis.

Conclusions: This study has shown that some patients are prone to episodes of hypoxemia during the course of normal daily activity or daytime sleep that would usually go unrecognized and that more effective management of supplemental oxygen treatment may be possible with continual unobtrusive monitoring.     

A comparative study of automatic techniques for ocular artifact reduction in spontaneous EEG signals based on clinical target variables: a simulation case

Eye movement artifacts represent a critical issue for quantitative electroencephalography (EEG) analysis and a number of mathematical approaches have been proposed to reduce their contribution in EEG recordings. The aim of this paper was to objectively and quantitatively evaluate the performance of ocular filtering methods with respect to spectral target variables widely used in clinical and functional EEG studies. In particular the following methods were applied: regression analysis and some blind source separation (BSS) techniques based on second-order statistics (PCA, AMUSE and SOBI) and on higher-order statistics (JADE, INFOMAX and FASTICA). Considering blind source decomposition methods, a completely automatic procedure of BSS based on logical rules related to spectral and topographical information was proposed in order to identify the components related to ocular interference. The automatic procedure was applied in different montages of simulated EEG and electrooculography (EOG) recordings: a full montage with 19 EEG and 2 EOG channels, a reduced one with only 6 EEG leads and a third one where EOG channels were not available. Time and frequency results in all of them indicated that AMUSE and SOBI algorithms preserved and recovered more brain activity than the other methods mainly at anterior regions. In the case of full montage: (i) errors were lower than 5% for all spectral variables at anterior sites; and (ii) the highest improvement in the signal-to-artifact (SAR) ratio was obtained up to 40dB at these anterior sites. Finally, we concluded that second-order BSS-based algorithms (AMUSE and SOBI) provided an effective technique for eye movement removal even when EOG recordings were not available or when data length was short.     

Adaptive spatio-temporal filtering of disturbed ECGs: a multi-channel approach to heartbeat detection in smart clothing

Intermittent disturbances are common in ECG signals recorded with smart clothing: this is mainly because of displacement of the electrodes over the skin. We evaluated a novel adaptive method for spatio-temporal filtering for heartbeat detection in noisy multi-channel ECGs including short signal interruptions in single channels. Using multi-channel database recordings (12-channel ECGs from 10 healthy subjects), the results showed that multi-channel spatio-temporal filtering outperformed regular independent component analysis. We also recorded seven channels of ECG using a T-shirt with textile electrodes. Ten healthy subjects performed different sequences during a 10-min recording: resting, standing, flexing breast muscles, walking and pushups. Using adaptive multi-channel filtering, the sensitivity and precision was above 97% in nine subjects. Adaptive multi-channel spatio-temporal filtering can be used to detect heartbeats in ECGs with high noise levels. One application is heartbeat detection in noisy ECG recordings obtained by integrated textile electrodes in smart clothing.     

Quantitative evaluation of artifact removal in real magnetoencephalogram signals with blind source separation

The magnetoencephalogram (MEG) is contaminated with undesired signals, which are called artifacts. Some of the most important ones are the cardiac and the ocular artifacts (CA and OA, respectively), and the power line noise (PLN). Blind source separation (BSS) has been used to reduce the influence of the artifacts in the data. There is a plethora of BSS-based artifact removal approaches, but few comparative analyses. In this study, MEG background activity from 26 subjects was processed with five widespread BSS (AMUSE, SOBI, JADE, extended Infomax, and FastICA) and one constrained BSS (cBSS) techniques. Then, the ability of several combinations of BSS algorithm, epoch length, and artifact detection metric to automatically reduce the CA, OA, and PLN were quantified with objective criteria. The results pinpointed to cBSS as a very suitable approach to remove the CA. Additionally, a combination of AMUSE or SOBI and artifact detection metrics based on entropy or power criteria decreased the OA. Finally, the PLN was reduced by means of a spectral metric. These findings confirm the utility of BSS to help in the artifact removal for MEG background activity.     

一种独立成分分析的鲁棒算法及其在脑磁图数据分析中的应用

独立成分分析是一种新的信号处理统计方法。被广泛用于各个领域。在信号分析中面临的难题是：源信号的不同特性（既包括超高斯信号又包括亚高斯信号）；未知的独立源数目；传感器信号受到较大的加性噪声污染。针对以上难题，本文提出了一种独立成分分析的鲁棒算法。该方法先对观测数据作预处理，将包含噪声的高维传感器观测信号降维分解到信号子空间和噪声子空间。利用交叉验证法估计出独立源的数目（解决了独立成分分析本身不能确定源数目的缺陷）；然后利用快速稳定的FastICA算法分离独立成分。通过人工合成的数据和实际的脑磁图数据分析。验证了这种方法的功效。     

Maternal and foetal ECG separation using blind source separation methods

The separation of the maternal and foetal electrocardiograms(ECGs) from skin electrodes located on the mother's body maybe modelled as a blind source separation (BSS) problem. Thisconsists in the reconstruction of a set of unknown mutuallyindependent source signals from the sole knowledge of anotherset of linear mixtures of the sources, where the mixture patternis also unknown. Three BSS methods based on cumulants are considered:principal-component analysis (PCA), higher-order singular-valuedecomposition (HOSVD), and higher-order eigenvalue decomposition(HOEVD). All these methods are applied to the foetal-ECG extractionproblem by using real ECG data. The last two methods appearto provide a more satisfactory separation than the first method,with HOEVD offering slightly better results.     

基于多区域分析的非接触式热红外视频心率检测方法

本研究提出了一种基于面部多区域分析的非接触式热红外视频心率检测方法.首先,确定面部3个感兴趣区域(region of interests,ROIs),构建像素均值时间序列.其次,对3个ROIs采用独立成分分析和多变量经验分解算法,分别提取包含心率信息的独立分量和本征模态函数,通过功率谱分析确定最佳独立分量和最佳本征模态...     

Extraction of cognitive activity-related waveforms from functional near-infrared spectroscopy signals

We address the problem of prototypical waveform extraction in cognitive experiments using functional near-infrared spectroscopy (fNIRS) signals. These waveform responses are evoked with visual stimuli provided in an oddball type experimental protocol. As the statistical signal-processing tool, we consider the linear signal space representation paradigm and use independent component analysis (ICA). The assumptions underlying ICA is discussed in the light of the signal measurement and generation mechanisms in the brain. The ICA-based waveform extraction is validated based both on its conformance to the parametric brain hemodynamic response (BHR) model and to the coherent averaging technique. We assess the intra-subject and inter-subject waveform and parameter variability.     

应用独立分量分析去除体表肌电中的心电干扰

体表肌电特别是从躯干获得的体表肌电往往受到被测对象自身心电信号的严重干扰。本文利用一种基于独立分量分析(ICA)的去噪方法,去除体表肌电中的心电干扰。该方法将多通道体表肌电进行独立分量分解,并用高通滤波器处理所分解出的心电独立分量以尽可能地保留其中的肌电成分,进而将去除心电干扰后的所有独立分量反向投影回原始信号空间得到去噪后的信号。仿真信号的处理结果表明,当高通滤波器的截止频率为30Hz时,该方法在有效去除心电干扰的同时使体表肌电的保真度达到最大。同时讨论了将信号的峰度(Kurtosis)值作为自动判别心电分量和肌电分量的标准的可能性。     

基于独立成分分析和量子粒子群算法的胎儿心电提取

胎儿心电图(FECG)是反映胎儿心脏电生理活动的一项客观指标,获取的FECG受到母体心电图(MECG)的干扰,如何快捷、有效的提取FECG成为重要的研究课题。在非侵入方式下,FECG的提取算法中独立成分分析(ICA)算法被认为是效果最好的方法,但现有求解其分解矩阵的算法收敛性能都不太高。量子粒子群(QPSO)算法是一种收敛于全局的智能优化算法。因此,提出了一种结合QPSO的ICA方法。研究结果表明,与其他在非侵入方式下的主要提取算法相比,这种方法能更清晰准确地提取出有用信号,为胎儿的健康检测提供了更好的方法。     

Consistency of the blind source separation computed with five common algorithms for magnetoencephalogram background activity

Blind source separation (BSS) is widely used to analyse brain recordings like the magnetoencephalogram (MEG). However, few studies have compared different BSS decompositions of real brain data. Those comparisons were usually limited to specific applications. Therefore, we aimed at studying the consistency (i.e., similarity) of the decompositions estimated for real MEGs from 26 subjects using five widely used BSS algorithms (AMUSE, SOBI, JADE, extended-Infomax and FastICA) for five epoch lengths (10 s, 20 s, 40 s, 60 s and 90 s). A statistical criterion based on Factor Analysis was applied to calculate the number of components into which each epoch would be decomposed. Then, the BSS techniques were applied. The results indicate that the pair of algorithms ‘AMUSE–SOBI’, followed by ‘JADE–FastICA’, provided the most similar separations. On the other hand, the most dissimilar outcomes were computed with ‘AMUSE–JADE’ and ‘SOBI–JADE’. The BSS decompositions were more similar for longer epochs. Furthermore, additional analyses of synthetic signals supported the results of the real MEGs. Thus, when selecting BSS algorithms to explore brain signals, the techniques offering the most different decompositions, such as AMUSE and JADE, may be preferred to obtain complementary, or at least different, perspectives of the underlying components.     

Comprehensive evaluation of PCA-based finite element modelling of the human femur

Computed tomography (CT)-based finite element (FE) reconstructions describe shape and density distribution of bones. Both shape and density distribution, however, can vary a lot between individuals. Shape/density indexation (usually achieved by principal component analysis—PCA) can be used to synthesize realistic models, thus overcoming the shortage of CT-based models, and helping e.g. to study fracture determinants, or steer prostheses design. The aim of this study was to describe a PCA-based statistical modelling algorithm, and test it on a large CT-based population of femora, to see if it can accurately describe and reproduce bone shape, density distribution, and biomechanics.To this aim, 115 CT-datasets showing normal femoral anatomy were collected and characterized. Isotopological FE meshes were built. Shape and density indexation procedures were performed on the mesh database. The completeness of the database was evaluated through a convergence study. The accuracy in reconstructing bones not belonging to the indexation database was evaluated through (i) leave-one-out tests (ii) comparison of calculated vs. in-vitro measured strains.Fifty indexation modes for shape and 40 for density were necessary to achieve reconstruction errors below pixel size for shape, and below 10% for density. Similar errors for density, and slightly higher errors for shape were obtained when reconstructing bones not belonging to the database. The in-vitro strain prediction accuracy of the reconstructed FE models was comparable to state-of-the-art studies.In summary, the results indicate that the proposed statistical modelling tools are able to accurately describe a population of femora through finite element models.     

Tracking and detection of epileptiform activity in multichannel ictal EEG using signal subspace correlation of seizure source scalp topographies

Conventional methods for monitoring clinical (epileptiform) multichannel electroencephalogram (EEG) signals often involve morphological, spectral or time-frequency analysis on individual channels to determine waveform features for detecting and classifying ictal events (seizures) and inter-ictal spikes. Blind source separation (BSS) methods, such as independent component analysis (ICA), are increasingly being used in biomedical signal processing and EEG analysis for extracting a set of underlying source waveforms and sensor projections from multivariate time-series data, some of which reflect clinically relevant neurophysiological (epileptiform) activity. The work presents an alternative spatial approach to source tracking and detection in multichannel EEG that exploits prior knowledge of the spatial topographies of the sensor projections associated with the target sources. The target source sensor projections are obtained by ICA decomposition of data segments containing representative examples of target source activity, e.g. a seizure or ocular artifact. Source tracking and detection are then based on the subspace correlation between individual target sensor projections and the signal subspace over a moving window. Different window lengths and subspace correlation threshold criteria reflect transient or sustained target source activity. To study the behaviour and potential application of this spatial source tracking and detection approach, the method was used to detect (transient) ocular artifacts and (sustained) seizure activity in two segments of 25-channel EEG data recorded from one epilepsy patient on two separate occasions, with promising and intuitive results.