Isolation and minimization of head motion-induced signal variations in fMRI data using independent component analysis.

Task-related head movement during acquisition of fMRI data represents a serious confound for both motion correction and estimates of task-related activation. Cost functions implemented in most conventional motion-correction algorithms compare two volumes for similarity but fail to account for signal variability that is not due to motion (e.g., brain activation). We therefore recently proposed the theoretical basis for a novel method for fMRI motion correction, termed motion-corrected independent component analysis (MCICA), that allows for brain activation present in an fMRI time-series to be implicitly modeled and mitigates motion-induced signal changes without having to directly estimate the motion parameters (Liao et al., IEEE Transactions on Medical Imaging 2005;25:29-44). To explore the effects of non-movement-related signal changes on registration error, we performed several previously proposed test simulations (Freire et al., IEEE Transactions on Medical Imaging 2002;21:470-484) to evaluate the performance of MCICA and compare it with the conventional square-of-difference-based measures such as LS-SPM and LS-AIR. We demonstrate that for both simulated data and real fMRI images, the proposed MCICA method performs favorably. Specifically, in simulations MCICA was more robust to the addition of simulated activation, and did not lead to the detection of false activations after correction for simulated task-correlated motion. With actual data from a motor fMRI experiment, the time course of the derived continually task-related ICA component became more correlated with the underlying behavioral task after preprocessing with MCICA compared to other methods, and the associated activation map was more clustered in the primary motor and supplementary motor cortices without spurious activation at the brain edge. We conclude that assessing the statistical properties of a motion-corrupted volume in relation to other volumes in the series, as is done with MCICA, is an accurate means of differentiating between motion-induced signal changes and other sources of variability in fMRI data.     

基于ICA的右侧颞叶癫痫患者警觉相关脑网络的静息态功能磁共振研究

目的 联合运用行为学测试及静息态功能磁共振成像(rs-fMRI)方法,探讨右侧颞叶癫痫(rTLE)患者警觉相关脑网络特点及其损害.方法 对20例rTLE患者和18例对照行注意网络测试(ANT),比较其警觉网络效率.随后行rs-fMRI扫描,运用基于稳定的群组独立成分分析软件(MICA)提取警觉相关脑网络,并行组内分析和组间比较.提取rTLE患者组差异脑区功能连接Z值,与警觉网络效率行相关分析.结果 rTLE患者与对照组ANT测试警觉网络效率差异无统计学意义.rTLE患者组与对照组存在类似的警觉相关脑网络,网络内功能连接减弱的脑区为右顶下小叶、角回,功能连接增强的脑区为右额下回岛盖部、中央沟盖、额中回、背外侧额上回、楔叶、枕上回等脑区.右楔叶功能连接Z值与警觉网络效率相关.结论 rTLE患者存在警觉相关脑网络损害,可通过增强其他脑区功能连接进行代偿.     

Real-time display of artifact-free electroencephalography during functional magnetic resonance imaging and magnetic resonance spectroscopy in an animal model of epilepsy.

Simultaneous recording of electroencephalogram (EEG) and functional MRI (fMRI) or MR spectroscopy (MRS) can provide further insight into our understanding of the underlying mechanisms of neurologic disorders. Current technology for simultaneous EEG and MRI recording is limited by extensive postacquisition processing of the data. Real-time display of artifact-free EEG recording during fMRI/MRS studies is essential in studies that involve epilepsy to ensure that they address specific EEG features such as epileptic spikes or seizures. By optimizing the EEG recording equipment to maximize the common mode rejection ratio of its amplifiers, a unique EEG system was designed and tested that allowed real-time display of the artifact-free EEG during fMRI/MRS in an animal model of epilepsy. Spike recordings were optimized by suppression of the background EEG activity using fast-acting and easily controlled inhalational anesthesia. Artifact suppression efficiency of 70-100% was achieved following direct subtraction of referentially recorded filtered EEG tracings from active electrodes, which were located in close proximity to each other (over homologous occipital cortices) and a reference electrode. Two independent postacquisition processing tools, independent component analysis and direct subtraction of unfiltered digital EEG data in MATLAB, were used to verify the accuracy of real-time EEG display.     

Estimation of respiration-induced noise fluctuations from undersampled multislice fMRI data.

Functional MRI time series data are known to be contaminated by highly structured noise due to physiological fluctuations. Significant components of this noise are at frequencies greater than those critically sampled in standard multislice imaging protocols and are therefore aliased into the activation spectrum, compromising the estimation of functional activations and the determination of their significance. However, in this work it is demonstrated that unaliased noise information is available in multislice data, and can be used to estimate and reduce noise due to high-frequency respiratory-related fluctuations. Magn Reson Med 45:635-644, 2001. Published 2001 Wiley-Liss, Inc.     

Independent component analysis of fMRI data in the complex domain.

In BOLD fMRI a series of MR images is acquired and examined for task-related amplitude changes. These functional changes are small, so it is important to maximize detection efficiency. Virtually all fMRI processing strategies utilize magnitude information and ignore the phase, resulting in an unnecessary loss of efficiency. As the optimum way to model the phase information is not clear, a flexible modeling technique is useful. To analyze complex data sets, independent component analysis (ICA), a data-driven approach, is proposed. In ICA, the data are modeled as spatially independent components multiplied by their respective time-courses. There are thus three possible approaches: 1) the time-courses can be complex-valued, 2) the images can be complex-valued, or 3) both the time-courses and the images can be complex-valued. These analytic approaches are applied to data from a visual stimulation paradigm, and results from three complex analysis models are presented and compared with magnitude-only results. Using the criterion of the number of contiguous activated voxels at a given threshold, an average of 12-23% more voxels are detected by complex-valued ICA estimation at a threshold of /Z/ > 2.5. Additionally, preliminary results from the complex models reveal a phase modulation similar to the magnitude time-course in some voxels, and oppositely modulated in other voxels.     

Higher‐order contrast functions improve performance of independent component analysis of fMRI data

Purpose

To evaluate the performance of different contrast functions used in Independent Component Analysis (ICA) of functional magnetic resonance imaging (fMRI) data at low signal‐to‐noise ratio (SNR), present in fMRI paradigms such as resting‐state acquisitions.

Materials and Methods

Metrics were defined to estimate both the accuracy and robustness of contrast functions under varying source distributions. Simulations were performed to compare the performance of lower‐order (such as ln cosh) to higher‐order (such as kurtosis) contrast functions using Laplacian source distributions corrupted with Gaussian noise. The ln cosh and kurtosis contrast functions were also compared using resting‐state fMRI data from 10 normal adult volunteers.

Results

Higher‐order contrast functions provided superior performance compared to lower‐order contrast functions in the evaluation of metrics and via the simulations in the presence of a significant amount of noise. The performance of kurtosis was not statistically significantly different from that of a theoretically optimized contrast function. The choice of contrast function was found to result in substantial (R < 0.9) differences in 40% of the components found from the resting‐state fMRI data.

Conclusion

The use of higher‐order contrast functions, such as kurtosis, may provide superior performance in ICA analysis of fMRI data with low SNR. J. Magn. Reson. Imaging 2009;29:242–249. © 2008 Wiley‐Liss, Inc.     

Improved temporal clustering analysis method for detecting multiple response peaks in fMRI

PURPOSE: To develop an improved temporal clustering analysis (TCA) method for detecting multiple active peaks by running the method once. MATERIALS AND METHODS: Two cases of simulation data and a set of actual fMRI data from nine subjects were used to compare the traditional TCA method with the new method, termed extremum TCA (ETCA). The first case of simulation data simulated event-related activation and block activation in one cerebral area, and the second case simulated event-related activation and block activation in two cerebral areas. An in vivo visual stimulating experiment was performed on a 1.5T MR scanner. All imaging data were processed using both traditional TCA and the new method. RESULTS: The results of both the simulated and actual fMRI data show that the new method is more sensitive and exact than traditional TCA in detecting multiple response peaks. CONCLUSION: The new method is effective in detecting multiple activations even when the timing and location of the brain activation are completely unknown.     

Motion Artifacts in fMRI: Comparison of 2DFT with PR and Spiral Scan Methods

Activation signals based on BOLD contrast changes consequent to neuronal stimulation typically produce cortical intensity differences of < 10% at 1.5T. Hemodynamically driven pulsation of the brain can cause highly pulsatile phase shifts, which in turn result in motion artifacts whose intensity is larger than the activation signals in 2DFT scan methods. This paper presents a theoretical and experimental comparison of the magnitude of such artifacts for 2DFT and two other methods using non-Cartesian k-space trajectories. It is shown that artifacts increase with TR for 2DFT methods, and that projection reconstruction (PR) and spiral methods have significantly reduced artifact intensities, because these trajectories collect low spatial frequencies with every view. The spiral technique is found to be superior in terms of efficiency and motion insensitivity.     

Event-related fMRI technique for auditory processing with hemodynamics unrelated to acoustic gradient noise.

Acoustic noise from the imaging gradients presents a major difficulty in functional MRI (fMRI) studies of auditory cortical function. For studies involving hearing-impaired pediatric subjects, the auditory stimuli should be presented during completely silent gradient intervals. In addition, the scan time is limited by constraints involving subject motion and subject compliance. A novel event-related method for conducting fMRI studies of auditory function is proposed. Auditory stimuli are presented during completely silent gradient intervals, but using a variable TR. A general nonlinear model (GNLM) is proposed as a postprocessing methodology for the data. The technique increases the flexibility of the experimental design, with minimal loss of sensitivity compared to standard fMRI acquisition techniques, and may therefore be useful for fMRI studies of auditory function in hearing-impaired pediatric subjects.     

A parallel approach to STAP implementation for fMRI data

PURPOSE: To exploit the capabilities of parallel processing in applying the space-time adaptive processing (STAP) algorithm, previously explored on a small scale for functional magnetic resonance imaging (fMRI) applications, to conventional size fMRI data sets. MATERIALS AND METHODS: STAP is a two-dimensional filter that is able to locate fMRI activations in both space and frequency. It is applied here for the construction of brain activation maps in fMRI using Visual Age C, incorporating Engineering and Scientific Subroutine Library (ESSL) functions, compiled in 64-bit, and executed on an IBM SP supercomputer. RESULTS: Computer simulations incorporating actual MRI noise indicate that STAP, incorporated using the method of steepest descent, is feasible on conventional size data sets and exhibits an improvement in detecting activations over the more traditional cross correlation method of fMRI analysis when the response is unknown. CONCLUSION: STAP is feasible on traditional size fMRI data sets and useful in elucidating spatial and temporal connectivity.     

Independent component analysis‐based multifiber streamline tractography of the human brain

An independent component analysis‐based approach has been developed to estimate the orientations of two or three crossing fibers in a voxel to conduct human brain streamline tractography from diffusion data acquired along 25 gradient directions at a b‐value of 1000 sec/mm². The approach relies on unmixing signals from fibers mixed within, and spread over, a small cluster of 11 voxels. Simulation studies of diffusion data for two or three crossing fibers at signal‐to‐noise ratios of 15 and 30 suggest the accuracy to determine interfiber angles with independent component analysis is similar to that attained by a gaussian mixture and other multicompartmental models but at two orders of magnitude faster computational speed. Compared to previous multicompartmental models, independent component analysis visually shows good recovery of fiber orientations and tracts in the crossing region of commonly available orthogonal and 60° phantom diffusion datasets. A 3T MRI human studies show that in contrast to conventional streamline tractography and a multicompartment model, independent component analysis shows better recovery of the continuity of fronto‐occipital tracts and cingulum from regions where these tracts are mixed with corpus callosum and other pathways. Magn Reson Med, 2010. © 2010 Wiley‐Liss, Inc.     

Independent component analysis for automated decomposition of in vivo magnetic resonance spectra.

Fully automated methods for analyzing MR spectra would be of great benefit for clinical diagnosis, in particular for the extraction of relevant information from large databases for subsequent pattern recognition analysis. Independent component analysis (ICA) provides a means of decomposing signals into their constituent components. This work investigates the use of ICA for automatically extracting features from in vivo MR spectra. After its limits are assessed on artificial data, the method is applied to a set of brain tumor spectra. ICA automatically, and in an unsupervised fashion, decomposes the signals into interpretable components. Moreover, the spectral decomposition achieved by the ICA leads to the separation of some tissue types, which confirms the biochemical relevance of the components.     

Spinal fMRI during proprioceptive and tactile tasks in healthy subjects: activity detected using cross-correlation, general linear model and independent component analysis

Introduction  Functional MRI (fMRI) of the spinal cord is able to provide maps of neuronal activity. Spinal fMRI data have been analyzed in previous studies by calculating the cross-correlation (CC) between the stimulus and the time course of every voxel and, more recently, by using the general linear model (GLM). The aim of this study was to compare three different approaches (CC analysis, GLM and independent component analysis (ICA)) for analyzing fMRI scans of the cervical spinal cord. Methods  We analyzed spinal fMRI data from healthy subjects during a proprioceptive and a tactile stimulation by using two model-based approaches, i.e., CC analysis between the stimulus shape and the time course of every voxel, and the GLM. Moreover, we applied independent component analysis, a model-free approach which decomposes the data in a set of source signals. Results  All methods were able to detect cervical cord areas of activity corresponding to the expected regions of neuronal activations. Model-based approaches (CC and GLM) revealed similar patterns of activity. ICA could identify a component correlated to fMRI stimulation, although with a lower statistical threshold than model-based approaches, and many components, consistent across subjects, which are likely to be secondary to noise present in the data. Conclusions  Model-based approaches seem to be more robust for estimating task-related activity, whereas ICA seems to be useful for eliminating noise components from the data. Combined use of ICA and GLM might improve the reliability of spinal fMRI results.     

A STAP algorithm approach to fMRI: a simulation study

PURPOSE: To adapt the space-time adaptive processing (STAP) algorithm, previously developed in the field of sensor array processing and applied to radar signal processing, for use in construction of brain activation maps in functional magnetic resonance imaging (fMRI). MATERIALS AND METHODS: STAP is a two-dimensional filter in which both the spatial and temporal responses are controlled adaptively. It processes space-time data as a complete spatiotemporal set. Unlike presently used fMRI techniques, STAP locates activated regions both spatially and in frequency. RESULTS: Computer simulations incorporating actual MRI noise indicate that STAP exhibits a high degree of accuracy in detecting the small signal intensity changes inherent in fMRI. CONCLUSION: Because STAP processes space-time data as a single data matrix, it exhibits potential over currently available fMRI methods in providing a measure of the full spatiotemporal extent of a task-related activity.     

Automated processing for proton spectroscopic imaging using water reference deconvolution

Automated formation of MR spectroscopic images (MRSI) is necessary before routine application of these methods is possible for in vivo studies; however, this task is complicated by the presence of spatially dependent instrumental distortions and the complex nature of the MR spectrum. A data processing method is presented for completely automated formation of in vivo proton spectroscopic images, and applied for analysis of human brain metabolites. This procedure uses the water reference deconvolution method (G. A. Morris, J. Magn. Reson. 80,547(1988)) to correct for line shape distortions caused by instrumental and sample characteristics, followed by parametric spectral analysis. Results for automated image formation were found to compare favorably with operator dependent spectral integration methods. While the water reference deconvolution processing was found to provide good correction of spatially dependent resonance frequency shifts, it was found to be susceptible to errors for correction of line shape distortions. These occur due to differences between the water reference and the metabolite distributions.     

Quantitative analysis in clinical applications of brain MRI using independent component analysis coupled with support vector machine

Purpose:

To effectively perform quantification of brain normal tissues and pathologies simultaneously, independent component analysis (ICA) coupled with support vector machine (SVM) is investigated and evaluated for effective volumetric measurements of normal and lesion tissues using multispectral MR images.

Materials and Methods:

Synthetic and real MR data of normal brain and white matter lesion (WML) data were used to evaluate the accuracy and reproducibility of gray matter (GM), white matter (WM), and WML volume measurements by using the proposed ICA+SVM method to analyze three sets of MR images, T1‐weighted, T2‐weighted, and proton density/fluid‐attenuated inversion recovery images.

Results:

The Tanimoto indexes of GM/WM classification in the normal synthetic data calculated by the ICA+SVM method were 0.82/0.89 for data with 0% noise level. As for clinical MR data experiments, the ICA+SVM method clearly extracted the normal tissues and white matter hyperintensity lesions from the MR images, with low intra‐ and inter‐operator coefficient of variations.

Conclusion:

The experiments conducted provide evidence that the ICA+SVM method has shown promise and potential in applications to classification of normal and pathological tissues in brain MRI. J. Magn. Reson. Imaging 2010;32:24–34. © 2010 Wiley‐Liss, Inc.     

Spatially filtering functional magnetic resonance imaging data

When constructing MR images from acquired spatial frequency data, it can be beneficial to apply a low-pass filter to remove high frequency noise from the resulting images. This amounts to attenuating high spatial frequency fluctuations that can affect detected MR signal. A study is presented of spatially filtering MR data and possible ramifications on detecting regionally specific activation signal. It is shown that absolute activation levels are strongly dependent on the parameters of the filter used in image construction and that significance of an activation signal can be enhanced through appropriate filter selection. A comparison is made between spatially filtering MR image data and applying a Gaussian convolution kernel to statistical parametric maps.