基于独立分量分析的诱发电位信号噪声分离方法

诱发电位(EP)信号的检测与分析技术是临床医学诊断神经系统损伤及病变的重要手段之一。但是，从人体体表所得到的EP信号含有大量的噪声，最典型的噪声是人体自发产生的脑电图信号(EEG)。因此，为利用EP信号诊断神经系统的损伤和病变，需要从混合信号中去除EEG等噪声。独立分量分析(ICA)是一种新近发展起来的统计信号处理方法。本文把ICA方法应用于EP信号的噪声消除，并与传统的自适应滤波方法进行了比较。计算机模拟表明，采用ICA方法进行信号噪声分离的结果明显优于自适应滤波方法。     

Robust extraction of P300 using constrained ICA for BCI applications

P300 is a positive event-related potential used by P300-brain computer interfaces (BCIs) as a means of communication with external devices. One of the main requirements of any P300-based BCI is accuracy and time efficiency for P300 extraction and detection. Among many attempted techniques, independent component analysis (ICA) is currently the most popular P300 extraction technique. However, since ICA extracts multiple independent components (ICs), its use requires careful selection of ICs containing P300 responses, which limits the number of channels available for computational efficiency. Here, we propose a novel procedure for P300 extraction and detection using constrained independent component analysis (cICA) through which we can directly extract only P300-relevant ICs. We tested our procedure on two standard datasets collected from healthy and disabled subjects. We tested our procedure on these datasets and compared their respective performances with a conventional ICA-based procedure. Our results demonstrate that the cICA-based method was more reliable and less computationally expensive, and was able to achieve 97 and 91.6% accuracy in P300 detection from healthy and disabled subjects, respectively. In recognizing target characters and images, our approach achieved 95 and 90.25% success in healthy and disabled individuals, whereas use of ICA only achieved 83 and 72.25%, respectively. In terms of information transfer rate, our results indicate that the ICA-based procedure optimally performs with a limited number of channels (typically three), but with a higher number of available channels (>3), its performance deteriorates and the cICA-based one performs better.     

Applicability of independent component analysis on high-density microelectrode array recordings

Emerging complementary metal oxide semiconductor (CMOS)-based, high-density microelectrode array (HD-MEA) devices provide high spatial resolution at subcellular level and a large number of readout channels. These devices allow for simultaneous recording of extracellular activity of a large number of neurons with every neuron being detected by multiple electrodes. To analyze the recorded signals, spiking events have to be assigned to individual neurons, a process referred to as "spike sorting." For a set of observed signals, which constitute a linear mixture of a set of source signals, independent component (IC) analysis (ICA) can be used to demix blindly the data and extract the individual source signals. This technique offers great potential to alleviate the problem of spike sorting in HD-MEA recordings, as it represents an unsupervised method to separate the neuronal sources. The separated sources or ICs then constitute estimates of single-neuron signals, and threshold detection on the ICs yields the sorted spike times. However, it is unknown to what extent extracellular neuronal recordings meet the requirements of ICA. In this paper, we evaluate the applicability of ICA to spike sorting of HD-MEA recordings. The analysis of extracellular neuronal signals, recorded at high spatiotemporal resolution, reveals that the recorded data cannot be modeled as a purely linear mixture. As a consequence, ICA fails to separate completely the neuronal signals and cannot be used as a stand-alone method for spike sorting in HD-MEA recordings. We assessed the demixing performance of ICA using simulated data sets and found that the performance strongly depends on neuronal density and spike amplitude. Furthermore, we show how postprocessing techniques can be used to overcome the most severe limitations of ICA. In combination with these postprocessing techniques, ICA represents a viable method to facilitate rapid spike sorting of multidimensional neuronal recordings.     

A fully automatic ocular artifact suppression from EEG data using higher order statistics: Improved performance by wavelet analysis

Contamination of electroencephalographic (EEG) recordings with different kinds of artifacts is the main obstacle to the analysis of EEG data. Independent component analysis (ICA) is now a widely accepted tool for detection of artifacts in EEG data. One major challenge to artifact removal using ICA is the identification of the artifactual components. Although several strategies were proposed for automatically detecting the artifactual component during past several years, there is still little consensus on the criteria for automatic rejection of undesired components. In this paper we present a new identification procedure based on an efficient combination of independent component analysis (ICA), mutual information, and wavelet analysis for fully automatic ocular artifact suppression. The method does not require any offline training or determining the threshold levels for different markers. The results show that the proposed method could significantly enhance the ocular artifact detection and suppression. The results on 3105 4-s EEG epochs indicate that the artifact components can be identified with an accuracy of 97.8%, a sensitivity of 96.9%, and a specificity of 98.6%.     

Differentially variable component analysis: Identifying multiple evoked components using trial-to-trial variability

Electric potentials and magnetic fields generated by ensembles of synchronously active neurons, either spontaneously or in response to external stimuli, provide information essential to understanding the processes underlying cognitive and sensorimotor activity. Interpreting recordings of these potentials and fields is difficult because detectors record signals simultaneously generated by various regions throughout the brain. We introduce a novel approach to this problem, the differentially variable component analysis (dVCA) algorithm, which relies on trial-to-trial variability in response amplitude and latency to identify multiple components. Using simulations we demonstrate the importance of response variability to component identification, the robustness of dVCA to noise, and its ability to characterize single-trial data. We then compare the source-separation capabilities of dVCA with those of principal component analysis and independent component analysis. Finally, we apply dVCA to neural ensemble activity recorded from an awake, behaving macaque-demonstrating that dVCA is an important tool for identifying and characterizing multiple components in the single trial.     

Latency analysis of single auditory evoked M100 responses by spatio-temporal filtering

Appropriate spatial filtering followed by temporal filtering is well suited for the single-trial analysis of multi-channel magnetoencephalogram or electroencephalogram recordings. This is demonstrated by the results of a single-trial latency analysis obtained for auditory evoked M100 responses from nine subjects using two different stimulation frequencies. Spatial filters were derived automatically from the data via noise-adjusted principle component analysis, and single-trial latencies were estimated from the signal phase after complex bandpass filtering. For each of the two stimulation frequencies, estimated single-trial latencies were consistent with results obtained from a standard approach using averaged evoked responses. The quality of the estimated single-trial latencies was additionally assessed by their ability to separate between the two different stimulation frequencies. As a result, more than 80% of the single trials can be classified correctly by their estimated latencies.     

Theta rhythmicities following expected visual and auditory targets.

Evoked (EPs) as well as event-related potentials (ERPs) were recorded from two groups of 10 healthy, voluntary subjects in auditory and visual modalities. For ERP recordings 'the omitted stimulus paradigm' was employed, in which the subjects were expected to mark mentally the onset time (time prediction task) of the omitted stimulus (target). The standard auditory (AEP) and visual (VEP) evoked potentials and auditory and visual ERPs to the preceding stimuli of the omitted ones were analyzed in time and frequency domains. In the time domain the time prediction task induced increases of the amplitudes of waves existing in standard EPs; however, an additional wave or component could not be detected. Analysis of amplitude frequency characteristics (AFCs) revealed, however, selective, significant increases of the theta (3-6 Hz) frequency components of the responses concerned. These theta increases were especially evident in the frontal and parietal recording sites. Our findings suggest an association between the theta frequency components of transient evoked responses, the association areas of the brain and cognitive performance. The neurophysiological basis of scalp recorded ERPs are discussed in relation to the findings of animal studies with EEG and single unit recordings from cortical and subcortical structures.     

Comparing ICA-based and Single-Trial Topographic ERP Analyses

Single-trial analysis of human electroencephalography (EEG) has been recently proposed for better understanding the contribution of individual subjects to a group-analyis effect as well as for investigating single-subject mechanisms. Independent Component Analysis (ICA) has been repeatedly applied to concatenated single-trial responses and at a single-subject level in order to extract those components that resemble activities of interest. More recently we have proposed a single-trial method based on topographic maps that determines which voltage configurations are reliably observed at the event-related potential (ERP) level taking advantage of repetitions across trials. Here, we investigated the correspondence between the maps obtained by ICA versus the topographies that we obtained by the single-trial clustering algorithm that best explained the variance of the ERP. To do this, we used exemplar data provided from the EEGLAB website that are based on a dataset from a visual target detection task. We show there to be robust correpondence both at the level of the activation time courses and at the level of voltage configurations of a subset of relevant maps. We additionally show the estimated inverse solution (based on low-resolution electromagnetic tomography) of two corresponding maps occurring at approximately 300 ms post-stimulus onset, as estimated by the two aforementioned approaches. The spatial distribution of the estimated sources significantly correlated and had in common a right parietal activation within Brodmann’s Area (BA) 40. Despite their differences in terms of theoretical bases, the consistency between the results of these two approaches shows that their underlying assumptions are indeed compatible.     

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.     

Denoising of HD-sEMG signals using canonical correlation analysis

High-density surface electromyography (HD-sEMG) is a recent technique that overcomes the limitations of monopolar and bipolar sEMG recordings and enables the collection of physiological and topographical informations concerning muscle activation. However, HD-sEMG channels are usually contaminated by noise in an heterogeneous manner. The sources of noise are mainly power line interference (PLI), white Gaussian noise (WGN) and motion artifacts (MA). The spectral components of these disruptive signals overlap with the sEMG spectrum which makes classical filtering techniques non effective, especially during low contraction level recordings. In this study, we propose to denoise HD-sEMG recordings at 20 % of the maximum voluntary contraction by using a second-order blind source separation technique, named canonical component analysis (CCA). For this purpose, a specific and automatic canonical component selection, using noise ratio thresholding, and a channel selection procedure for the selective version (sCCA) are proposed. Results obtained from the application of the proposed methods (CCA and sCCA) on realistic simulated data demonstrated the ability of the proposed approach to retrieve the original HD-sEMG signals, by suppressing the PLI and WGN components, with high accuracy (for five different simulated noise dispersions using the same anatomy). Afterward, the proposed algorithms are employed to denoise experimental HD-sEMG signals from five healthy subjects during biceps brachii contractions following an isometric protocol. Obtained results showed that PLI and WGN components could be successfully removed, which enhances considerably the SNR of the channels with low SNR and thereby increases the mean SNR value among the grid. Moreover, the MA component is often isolated on specific estimated sources but requires additional signal processing for a total removal. In addition, comparative study with independent component analysis, CCA-wavelet and CCA-empirical mode decomposition (EMD) proved a higher efficiency of the presented method over existing denoising techniques and demonstrated pointless a second filtering stage for denoising HD-sEMG recordings at this contraction level.     

Removal of the skin blood flow artifact in functional near-infrared spectroscopic imaging data through independent component analysis

We investigate whether the functional near-infrared spectroscopic (fNIRS) signal includes a signal from the changing skin blood flow. During a locomotor task on a treadmill, changes in the hemodynamic response in the front-parietal area of healthy human subjects are simultaneously recorded using an fNIRS imaging system and a laser Doppler tissue blood flow meter. Independent component analysis (ICA) for fNIRS signals is performed. The skin blood flow changes during locomotor tasks on a treadmill. The activated spatial distribution of one of the components separated by ICA reveals an overall increase in fNIRS channels. To evaluate the uniformity of the activated spatial distribution, we define a new statistical value-the coefficient of spatial uniformity (CSU). The CSU value is a highly discriminating value (e.g., 2.82) compared with values of other components (e.g., 1.41, 1.10, 0.96, 0.61, and 0.58). In addition, the independent component signal corresponding to the activated spatial distribution is similar to changes in skin blood flow measured with the laser Doppler tissue blood flow meter. The coefficient of correlation indicates strong correlation. Localized activation areas around the premotor and medial somatosensory cortices are shown more clearly by eliminating the extracted component.     

独立分量分析的拟牛顿迭代算法及其在视觉诱发电位特征提取中的应用

少次叠加平均处理后的视觉诱发电位（VEP）中仍含有一定的背景噪声．对其进行进一步的提取与处理有重要的实用价值。独立分量分析（ICA）能够从混合信号中分离出最独立的成分，有效抑制噪声。本文尝试采用ICA的拟牛顿迭代算法进行VEP特征提取，介绍该方法的原理、实验和结果，并与采用牛顿迭代准则的快速独立孕量分析（Fast ICA）算法进行了比较。结果表明，基于拟牛顿法的ICA可以有效增强信号，从少次叠加平均的结果中提取出易于辨识的VEP的P300信号，具有较高的应用价值。     

Extracting multisource brain activity from a single electromagnetic channel

This paper develops a methodology for the extraction of multisource brain activity using only single channel recordings of electromagnetic (EM) brain signals. Measured electroencephalogram (EEG) and magnetoencephalogram (MEG) signals are used to demonstrate the utility of the method on extracting multisource activity from a single channel recording. At the heart of the method is dynamical embedding (DE) where first an appropriate embedding matrix is constructed out of a series of delay vectors from the measured signal. The embedding matrix contains the information we require, but in a mixed form which therefore needs to be deconstructed. In particular, we demonstrate how one form of independent component analysis (ICA) performed on the embedding matrix can deconstruct the single channel recording into its underlying informative components. The components are treated as a convenient expansion basis and subjective methods are then used to identify components of interest relevant to the application. The framework has been applied to single channels of both EEG and MEG recordings and is shown to isolate multiple sources of activity which includes: (i) artifactual components such as ocular, electrocardiographic and electrode artefact, (ii) seizure components in epileptic EEG recordings, and (iii) theta band, tumour related, activity in MEG recordings. The results are intuitive and meaningful in a neurophysiological setting.     

Validation of ICA as a tool to remove eye movement artifacts from EEG/ERP

Eye movement artifacts in electroencephalogram (EEG) recordings can greatly distort grand mean event‐related potential (ERP) waveforms. Different techniques have been suggested to remove these artifacts prior to ERP analysis. Independent component analysis (ICA) is suggested as an alternative method to “filter” eye movement artifacts out of the EEG, preserving the brain activity of interest and preserving all trials. However, the identification of artifact components is not always straightforward. Here, we compared eye movement artifact removal by ICA compiled on 10 s of EEG, on eye movement epochs, or on the complete EEG recording to the removal of eye movement artifacts by rejecting trials or by the Gratton and Coles method. ICA performed as well as the Gratton and Coles method. By selecting only eye movement epochs for ICA compilation, we were able to facilitate the identification of components representing eye movement artifacts.     

A new approach to spike sorting for multi-neuronal activities recorded with a tetrode--how ICA can be practical

Multi-neuronal recording with a tetrode is a powerful technique to reveal neuronal interactions in local circuits. However, it is difficult to detect precise spike timings among closely neighboring neurons because the spike waveforms of individual neurons overlap on the electrode when more than two neurons fire simultaneously. In addition, the spike waveforms of single neurons, especially in the presence of complex spikes, are often non-stationary. These problems limit the ability of ordinary spike sorting to sort multi-neuronal activities recorded using tetrodes into their single-neuron components. Though sorting with independent component analysis (ICA) can solve these problems, it has one serious limitation that the number of separated neurons must be less than the number of electrodes. Using a combination of ICA and the efficiency of ordinary spike sorting technique (k-means clustering), we developed an automatic procedure to solve the spike-overlapping and the non-stationarity problems with no limitation on the number of separated neurons. The results for the procedure applied to real multi-neuronal data demonstrated that some outliers which may be assigned to distinct clusters if ordinary spike-sorting methods were used can be identified as overlapping spikes, and that there are functional connections between a putative pyramidal neuron and its putative dendrite. These findings suggest that the combination of ICA and k-means clustering can provide insights into the precise nature of functional circuits among neurons, i.e. cell assemblies.     

Automatic sorting for multi-neuronal activity recorded with tetrodes in the presence of overlapping spikes

Multi-neuronal recording is a powerful electrophysiological technique that has revealed much of what is known about the neuronal interactions in the brain. However, it is difficult to detect precise spike timings, especially synchronized simultaneous firings, among closely neighboring neurons recorded by one common electrode because spike waveforms overlap on the electrode when two or more neurons fire simultaneously. In addition, the non-Gaussian variability (nonstationarity) of spike waveforms, typically seen in the presence of so-called complex spikes, limits the ability to sort multi-neuronal activities into their single-neuron components. Because of these problems, the ordinary spike-sorting techniques often give inaccurate results. Our previous study has shown that independent component analysis (ICA) can solve these problems and separate single-neuron components from multi-neuronal recordings. The ICA has, however, one serious limitation that the number of separated neurons must be less than the number of electrodes. The present study combines the ICA and the efficiency of the ordinary spike-sorting technique (k-means clustering) to solve the spike-overlapping and the nonstationarity problems with no limitation on the number of single neurons to be separated. First, multi-neuronal activities are sorted into an overly large number of clusters by k-means clustering. Second, the sorted clusters are decomposed by ICA. Third, the decomposed clusters are progressively aggregated into a minimal set of putative single neurons based on similarities of basis vectors estimated by ICA. We applied the present procedure to multi-neuronal waveforms recorded with tetrodes composed of four microwires in the prefrontal cortex of awake behaving monkeys. The results demonstrate that there are functional connections among neighboring pyramidal neurons, some of which fire in a precise simultaneous manner and that precisely time-locked monosynaptic connections are working between neighboring pyramidal neurons and interneurons. Detection of these phenomena suggests that the present procedure can sort multi-neuronal activities, which include overlapping spikes and realistic non-Gaussian variability of spike waveforms, into their single-neuron components. We processed several types of synthesized data sets in this procedure and confirmed that the procedure was highly reliable and stable. The present method provides insights into the local circuit bases of excitatory and inhibitory interactions among neighboring neurons.     

基于独立分量分析的脑电噪声消除

作为一种新的多元统计处理方法，独立分量分析(ICA)是解决盲源分离(BSS)问题的一个有效手段。在简要分析ICA理论及其算法的基础上，提出将其应用到脑电中的眼电伪迹的去除任务。实际采集的生理信号大多由相互独立的成分线性迭加而成，符合ICA要求源信号统计独立的基本假设。与传统方法相比，ICA这种空间滤波器不受信号频谱混迭的限制，消噪的同时能对有用信号的细节成分做到很好的保留，很大程度上弥补了时频域方法的不足。此外解混矩阵的逆可以用来反映独立源的空间分布模式，具有重要的生理意义。     

A Nonparametric Statistical Evaluation of Changes in Evoked Potentials to Different Stimuli

An experiment yielding the evoked electroencephalographic potentials (EPs) to seven different stimuli was conducted on a population of 13 female and 7 male normal subjects. Patterned-visual, auditory and combined (bimodal) stimuli were used. One stimulus case required subject responses. Calculations of the first four statistical sample moments at the EP time points indicated nongaussian amplitude distributions. A nonparametric statistical test, the Mann-Whitney (U-test, was used to den it the significant point (p <.01) differences among the EPs to the visual, auditory and bimodal stimuli. The most extensive differences were obtained from EPs to stimulus pairs requiring subject response (identification). Differences among the EPs to different visual patterns existed to a lesser extent than to stimuli of different sensory modalities. In general, a change in the physical parameter of the stimulus induced changes in the earlier components, while the response requirement induced changes in the late components of the EP. No significant differences were found in EPs to the same stimuli between recording locations with hemispheric symmetry.     

Influence of task difficulty on the features of event-related potential during visual oddball task

We present a comprehensive analysis of the change in event-related potential (ERP) due to task difficulty during a visual oddball task. Specifically, we investigated the inter-subject difference in difficulty-related change of ERP patterns using single-trial ERP analysis focusing on P300 and P2 components. ERPs were recorded and analyzed from 14 subjects while performing a visual oddball task with two difficulty levels. After extracting independent components (ICs) from single-trial ERPs, the averaged ERPs were used to identify which ICs originated from major ERP components. The ERP components were estimated from single-trial waveforms by back-projecting relevant ICs onto scalp electrodes after removing all other ICs; thus, the comparison of ERP components could be performed for each subject. The averaged P300 amplitude was smaller and latency was larger for the more difficult task, and this tendency was also observed for single-trial ERP analysis within each subject. P2 amplitude increased for the hard task for both group and individual analyses, suggesting that the P2 may be interpreted as a manifestation of task relevance evaluation or response generation. The P2 amplitude and latency were more notably correlated with response time for the more difficult task.