Strategies for reducing large fMRI data sets for independent component analysis

In independent component analysis (ICA), principal component analysis (PCA) is generally used to reduce the raw data to a few principal components (PCs) through eigenvector decomposition (EVD) on the data covariance matrix. Although this works for spatial ICA (sICA) on moderately sized fMRI data, it is intractable for temporal ICA (tICA), since typical fMRI data have a high spatial dimension, resulting in an unmanageable data covariance matrix. To solve this problem, two practical data reduction methods are presented in this paper. The first solution is to calculate the PCs of tICA from the PCs of sICA. This approach works well for moderately sized fMRI data; however, it is highly computationally intensive, even intractable, when the number of scans increases. The second solution proposed is to perform PCA decomposition via a cascade recursive least squared (CRLS) network, which provides a uniform data reduction solution for both sICA and tICA. Without the need to calculate the covariance matrix, CRLS extracts PCs directly from the raw data, and the PC extraction can be terminated after computing an arbitrary number of PCs without the need to estimate the whole set of PCs. Moreover, when the whole data set becomes too large to be loaded into the machine memory, CRLS-PCA can save data retrieval time by reading the data once, while the conventional PCA requires numerous data retrieval steps for both covariance matrix calculation and PC extractions. Real fMRI data were used to evaluate the PC extraction precision, computational expense, and memory usage of the presented methods.     

Performance of blind source separation algorithms for fMRI analysis using a group ICA method

Independent component analysis (ICA) is a popular blind source separation technique that has proven to be promising for the analysis of functional magnetic resonance imaging (fMRI) data. A number of ICA approaches have been used for fMRI data analysis, and even more ICA algorithms exist; however, the impact of using different algorithms on the results is largely unexplored. In this paper, we study the performance of four major classes of algorithms for spatial ICA, namely, information maximization, maximization of non-Gaussianity, joint diagonalization of cross-cumulant matrices and second-order correlation-based methods, when they are applied to fMRI data from subjects performing a visuo-motor task. We use a group ICA method to study variability among different ICA algorithms, and we propose several analysis techniques to evaluate their performance. We compare how different ICA algorithms estimate activations in expected neuronal areas. The results demonstrate that the ICA algorithms using higher-order statistical information prove to be quite consistent for fMRI data analysis. Infomax, FastICA and joint approximate diagonalization of eigenmatrices (JADE) all yield reliable results, with each having its strengths in specific areas. Eigenvalue decomposition (EVD), an algorithm using second-order statistics, does not perform reliably for fMRI data. Additionally, for iterative ICA algorithms, it is important to investigate the variability of estimates from different runs. We test the consistency of the iterative algorithms Infomax and FastICA by running the algorithm a number of times with different initializations, and we note that they yield consistent results over these multiple runs. Our results greatly improve our confidence in the consistency of ICA for fMRI data analysis.     

Optimizing ICA in fMRI using information on spatial regularities of the sources

Spatial independent component analysis (ICA) is a well-established technique for multivariate analysis of functional magnetic resonance imaging (fMRI) data. It blindly extracts spatiotemporal patterns of neural activity from functional measurements by seeking for sources that are maximally independent. Additional information on one or more sources (e.g., spatial regularity) is often available; however, it is not considered while looking for independent components. In the present work, we propose a new ICA algorithm based on the optimization of an objective function that accounts for both independence and other information on the sources or on the mixing model in a very general fashion. In particular, we apply this approach to fMRI data analysis and illustrate, by means of simulations, how inclusion of a spatial regularity term helps to recover the sources more effectively than with conventional ICA. The improvement is especially evident in high noise situations. Furthermore we employ the same approach on data sets from a complex mental imagery experiment, showing that consistency and physiological plausibility of relatively weak components are improved.     

Improved steered response power method for sound source localization based on principal eigenvector

Sound source localization is essential in many microphone arrays application, ranging from teleconferencing systems to artificial perception in a reverberant noisy environment. The steered response power (SRP) using the phase transform (SRP-PHAT) source localization algorithm has been proved robust, however, the performance of the SRP-PHAT algorithm degrades in highly reverberant noisy environment. Though the SRP-based maximum likelihood localizers are more robust than SRP-PHAT, they have the drawback of requiring noise variance to be estimated in a silent room. This paper presents an improved SRP-PHAT algorithm based on principal eigenvector. Sound source location is estimated from the principal eigenvector computed from the frequency-domain correlation matrix. Using both simulated and real data, we show that the proposed algorithm achieves higher source localization accuracy compared to the SRP-PHAT algorithm.     

海洋环境噪声场对称性分析及噪声消除方法

实际的海洋环境是非常复杂的,存在着海洋自噪声、舰船噪声、生物发声等,阵元接收到的噪声信号存在一定的相关性,此时基于传统阵列信号处理的目标方位估计方法的性能将变差,针对这一问题,提出了一种实部消除方法.首先从阵元接收环境噪声的物理机理出发,将圆环阵接收的噪声场分解为对称噪声场和非对称噪声场,并且研究发现对称噪声场只影响数据协方差矩阵的实部.然后通过消除协方差矩阵实部,达到消除对称噪声场的目的,提高信噪比,但是同时产生了虚假声源.针对虚假声源的问题,提出了基于优化算法重构协方差矩阵实部的方法,消除了虚假声源的影响.仿真分析与海试数据处理结果表明:该方法明显消除了对称噪声,提高了信噪比,改善了阵列信号处理算法的性能.实部消除方法易于实现,有一定的工程应用价值.     

Uncertainty estimation in simultaneous Bayesian tracking and environmental inversion

This paper develops a Bayesian approach for two related inverse problems: tracking an acoustic source when ocean environmental parameters are unknown, and determining environmental parameters using acoustic data from an unknown (moving) source. The formulation considers source and environmental parameters as unknown random variables constrained by noisy acoustic data and by prior information on parameter values (e.g., physical limits for environmental properties) and on inter-parameter relationships (limits on radial and vertical source speed). The goal is not simply to estimate parameter values, but to rigorously determine parameter uncertainty distributions, thereby quantifying the information content of the data/prior to resolve source and environmental parameters. Results are presented as marginal posterior probability densities (PPDs) for environmental parameters and joint marginal PPDs for source ranges and depths. Given the numerically intensive inversion, an efficient Markov-chain Monte Carlo importance-sampling approach is developed which combines Metropolis and heat-bath Gibbs' sampling, employs efficient proposal distributions based on a linearized PPD approximation, and considers nonunity sampling temperatures to ensure a complete parameter search. The approach is illustrated with two simulated examples representing tracking a quiet submerged source and geoacoustic inversion using noise from an unknown ship of opportunity. In both cases, source, seabed, and water-column parameters are unknown.     

Data error covariance in matched-field geoacoustic inversion

Many approaches to geoacoustic inversion are based implicitly on the assumptions that data errors are Gaussian-distributed and spatially uncorrelated (i.e., have a diagonal covariance matrix). However, the latter assumption is often not valid due to theory errors, and can lead to reduced accuracy for geoacoustic parameter estimates and underestimation of parameter uncertainties. This paper examines the effects of data error (residual) covariance in matched-field geoacoustic inversion. An inversion approach is developed based on a nonparametric method of estimating the full covariance matrix (including off-diagonal terms) from the data residuals and explicitly including this covariance in the misfit function. Qualitative and quantitative statistical tests for Gaussianity and for correlations in complex residuals are considered to validate the inversion results. The approach is illustrated for Bayesian geoacoustic inversion of broadband, vertical-array acoustic data measured in the Mediterranean Sea.     

Parametric time domain analysis of the multiple input/scalar output problem: The source identification problem

A multiple input/scalar output stationary time series identification problem is considered from a parametric model time domain point of view. Particular emphasis is on the source identification problem. Closed form formula estimates of the individual source power contributions are expressed in terms of sample correlations that are obtained from the observed input and output time series and from parametric models fitted to that data. The estimates of the noise power contributions are asymptotically jointly normally distributed. The mean values and covariance matrix of those estimates yield confidence interval estimates of the individual and joint power contributions.The motivation for developing a rational polynomial transfer function or ARMA model of the multi-input scalar output plus additive noise situation is given. A two correlated input/single output version of this model is considered for a Monte Carlo simulation study. Parametric ARMA and approximate AR models are fitted to the simulated data. The asymptotic normality, and the distribution of the mean and covariances of the source power contribution computed from the ARMA and AR models are appraised.Several facets of the relative performance of windowed periodogram and AR model spectral analysis are examined for the multiple input/scalar output identification problem. The points that are emphasized are that conventional windowed periodogram spectral analysis is subjective, not particularly satisfactory for the sharp spectral peak situation that is commonly encountered in vibration data analysis and very likely not as good as “objective” Akaike criterion order AR modelled spectral analysis.     

Error Covariance Matrix Estimation of Noisy and Dynamically Coupled Time Series

We estimate the covariance matrix of the errors in several dynamically coupled time series corrupted by measurement errors. We say that several scalar time series are dynamically coupled if they record the values of measurements of the state variables of the same smooth dynamical system. The estimation of the covariance matrix of the errors is made using a noise reduction algorithm that efficiently exploits the information contained jointly in the dynamically coupled noisy time series. The method is particularly powerful for short length time series with high uncertainties.     

Remote sensing of moving wide-band sources

We consider the methods of covariance matrix reconstruction for sensing an extended noise-like emitter which moves near the receiving antenna array and is modelled as a random longitudinal distribution of external elementary sources. Linearized maximum-likelihood equations for the desired covariance matrix are obtained in the cases of a priori known and unknown covariance matrix of the noise background. The variance of estimates for the elements of the covariance matrix characterizing the emitter is studied. Institute of Applied Physics of the Russian Academy of Sciences, Nizhny Novgorod, Russia. Translated from Izvestiya Vysshikh Uchebnykh Zavedenii, Radiofizika, Vol. 41, No. 9, pp. 1163–1176, September, 1998.     

Covariance-based approaches to aeroacoustic noise source analysis

In this paper, several covariance-based approaches are proposed for aeroacoustic noise source analysis under the assumptions of a single dominant source and all observers contaminated solely by uncorrelated noise. The Crame?r-Rao Bounds (CRB) of the unbiased source power estimates are also derived. The proposed methods are evaluated using both simulated data as well as data acquired from an airfoil trailing edge noise experiment in an open-jet aeroacoustic facility. The numerical examples show that the covariance-based algorithms significantly outperform an existing least-squares approach and provide accurate power estimates even under low signal-to-noise ratio (SNR) conditions. Furthermore, the mean-squared-errors (MSEs) of the so-obtained estimates are close to the corresponding CRB especially for a large number of data samples. The experimental results show that the power estimates of the proposed approaches are consistent with one another as long as the core analysis assumptions are obeyed.     

Comparison of two exploratory data analysis methods for fMRI: fuzzy clustering vs. principal component analysis

Exploratory data-driven methods such as Fuzzy clustering analysis (FCA) and Principal component analysis (PCA) may be considered as hypothesis-generating procedures that are complementary to the hypothesis-led statistical inferential methods in functional magnetic resonance imaging (fMRI). Here, a comparison between FCA and PCA is presented in a systematic fMRI study, with MR data acquired under the null condition, i.e., no activation, with different noise contributions and simulated, varying "activation." The contrast-to-noise (CNR) ratio ranged between 1-10. We found that if fMRI data are corrupted by scanner noise only, FCA and PCA show comparable performance. In the presence of other sources of signal variation (e.g., physiological noise), FCA outperforms PCA in the entire CNR range of interest in fMRI, particularly for low CNR values. The comparison method that we introduced may be used to assess other exploratory approaches such as independent component analysis or neural network-based techniques.     

The asymptotic spectrum of the EWMA covariance estimator

The exponentially weighted moving average (EWMA) covariance estimator is a standard estimator for financial time series, and its spectrum can be used for so-called random matrix filtering. Random matrix filtering using the spectrum of the sample covariance matrix is an established tool in finance and signal detection and the EWMA spectrum can be used analogously. In this paper, the asymptotic spectrum of the EWMA covariance estimator is calculated using the Mar?enko-Pastur theorem. Equations for the spectrum and the boundaries of the support of the spectrum are obtained and solved numerically. The spectrum is compared with covariance estimates using simulated i.i.d. data and log-returns from a subset of stocks from the S&P 500. The behaviour of the EWMA estimator in this limited empirical study is similar to the results in previous studies of sample covariance matrices. Correlations in the data are found to only affect a small part of the EWMA spectrum, suggesting that a large part may be filtered out.     

波达方向估计中特征空间的信源数估计方法

提出了特征空间法信源数估计方法,它将阵列信号的协方差估计值分别投影到信号的特征子空间和噪声的特征子空间。由于信号子空间与噪声子空间相互正交,易于由表征投影大小的判据值区分信号和噪声的贡献;本方法用的是M×M阶矩阵特征值分解,M为基元数,与波达方向估计用的相同,因此节省大量的计算量;它可以在实数空间中进行运算,进一步减少运算量。进行了数值计算,检验了判据值分布,以及在信源等功率、不等功率和空间相关色噪声等情况下特征空间法的性能。估计方法还用声纳数据进行了检验。所有这些结果均证明本估计方法性能优良。      

Validating the performance of one-time decomposition for fMRI analysis using ICA with automatic target generation process

Independent component analysis (ICA) has been proven to be effective for functional magnetic resonance imaging (fMRI) data analysis. However, ICA decomposition requires to optimize the unmixing matrix iteratively whose initial values are generated randomly. Thus the randomness of the initialization leads to different ICA decomposition results. Therefore, just one-time decomposition for fMRI data analysis is not usually reliable. Under this circumstance, several methods about repeated decompositions with ICA (RDICA) were proposed to reveal the stability of ICA decomposition. Although utilizing RDICA has achieved satisfying results in validating the performance of ICA decomposition, RDICA cost much computing time. To mitigate the problem, in this paper, we propose a method, named ATGP-ICA, to do the fMRI data analysis. This method generates fixed initial values with automatic target generation process (ATGP) instead of being produced randomly. We performed experimental tests on both hybrid data and fMRI data to indicate the effectiveness of the new method and made a performance comparison of the traditional one-time decomposition with ICA (ODICA), RDICA and ATGP-ICA. The proposed method demonstrated that it not only could eliminate the randomness of ICA decomposition, but also could save much computing time compared to RDICA. Furthermore, the ROC (Receiver Operating Characteristic) power analysis also denoted the better signal reconstruction performance of ATGP-ICA than that of RDICA.     

轴对称介质内辐射源项反演算法

结合Abel变换和离散坐标法,提出了一种基于CCD相机采集的单幅辐射图像重建轴对称发射-吸收介质内辐射源项分布的反演算法。通过在求解辐射正问题得到的准确值的基础上,添加随机噪声模拟试验测量数据,分析了网格数目、辐射源项分布形式、吸收系数和测量误差对算法反演精度的影响。测试结果表明:该算法对测量误差不敏感,在有测量误差的情况下也能够准确的重建介质内的辐射源项分布。     

Blind noisy image separation based on a new robust independent component analysis network

The separation of noisy image is a very exciting area of research, especially when no prior information is available about the noisy image. In this paper, we propose a robust independent component analysis(ICA) network for separation images contaminated with high-level additive noise or outliers. We reduce the power of additive noise by adding outlier rejection rule in ICA. Extensive computer simulations confirm robustness and the excellent performance of the resulting algorithms.     

Derivative temporal clustering analysis: detecting prolonged neuronal activity

Temporal clustering analysis (TCA) and independent component analysis (ICA) are promising data-driven techniques in functional magnetic resonance imaging (fMRI) experiments to obtain brain activation maps in conditions with unknown temporal information regarding the neuronal activity. Although comparable to ICA in detecting transient neuronal activities, TCA fails to detect prolonged plateau brain activations. To eliminate this pitfall, a novel derivative TCA (DTCA) method was introduced and its algorithms with different subtraction intervals were tested on simulated data with a pattern of prolonged plateau brain activation. It was found that the best performance of DTCA method in generating functional maps could be obtained if the subtraction interval is equal to or larger than the length of the rising time of the fMRI response. The DTCA method and its theoretical predication were further investigated and validated using in vivo fMRI data sets. By removing the limitations in the previous TCA, DTCA has shown its powerful capability in detecting prolonged plateau neuronal activities.     

Discussion on the choice of separated components in fMRI data analysis by spatial independent component analysis

By measuring the changes of magnetic resonance signals during a stimulation, the functional magnetic resonance imaging (fMRI) is able to localize the neural activation in the brain. In this report, we discuss the fMRI application of the spatial independent component analysis (spatial ICA), which maximizes statistical independence over spatial images. Included simulations show the possibility of the spatial ICA on discriminating asynchronous activations or different response patterns in an fMRI data set. An in vivo visual stimulation fMRI test was conducted, and the result shows a proper sum of the separated components as the final image is better than a single component, using fMRI data analysis by spatial ICA. Our result means that spatial ICA is a useful tool for the detection of different response activations and suggests that a proper sum of the separated independent components should be used for the imaging result of fMRI data processing.     

General second-order covariance of Gaussian maximum likelihood estimates applied to passive source localization in fluctuating waveguides

A method is provided for determining necessary conditions on sample size or signal to noise ratio (SNR) to obtain accurate parameter estimates from remote sensing measurements in fluctuating environments. These conditions are derived by expanding the bias and covariance of maximum likelihood estimates (MLEs) in inverse orders of sample size or SNR, where the first-order covariance term is the Cramer-Rao lower bound (CRLB). Necessary sample sizes or SNRs are determined by requiring that (i) the first-order bias and the second-order covariance are much smaller than the true parameter value and the CRLB, respectively, and (ii) the CRLB falls within desired error thresholds. An analytical expression is provided for the second-order covariance of MLEs obtained from general complex Gaussian data vectors, which can be used in many practical problems since (i) data distributions can often be assumed to be Gaussian by virtue of the central limit theorem, and (ii) it allows for both the mean and variance of the measurement to be functions of the estimation parameters. Here, conditions are derived to obtain accurate source localization estimates in a fluctuating ocean waveguide containing random internal waves, and the consequences of the loss of coherence on their accuracy are quantified.