Strategies for improving the detection of fMRI activation in trigeminal pathways with cardiac gating

Functional magnetic resonance imaging (fMRI) has become a powerful tool for studying the normal and diseased human brain. The application of fMRI in detecting neuronal signals in the trigeminal system, however, has been hindered by low detection sensitivity due to activation artifacts caused by cardiac pulse-induced brain and brainstem movement. A variety of cardiac gating techniques have been proposed to overcome this issue, typically by phase locking the sampling to a particular time point during each cardiac cycle. We sought to compare different cardiac gating strategies for trigeminal system fMRI. In the present study, we used tactile stimuli to elicit brainstem and thalamus activation and compared the fMRI results obtained without cardiac gating and with three different cardiac gating strategies: single-echo with TR of 3 or 9 heartbeats (HBs) and dual-echo T2*-mapping EPI (TR = 2 HBs, TE = 21/55 ms). The dual-echo T2* mapping and the single-echo with TR of 2 and 3 HBs cardiac-gated fMRI techniques both increased detection rate of fMRI activation in brainstem. Activation in the brainstem and the thalamus was best detected by cardiac-gated dual-echo EPI.     

Reducing inter-subject anatomical variation: Effect of normalization method on sensitivity of functional magnetic resonance imaging data analysis in auditory cortex and the superior temporal region

Conventional group analysis of functional MRI (fMRI) data usually involves spatial alignment of anatomy across participants by registering every brain image to an anatomical reference image. Due to the high degree of inter-subject anatomical variability, a low-resolution average anatomical model is typically used as the target template, and/or smoothing kernels are applied to the fMRI data to increase the overlap among subjects' image data. However, such smoothing can make it difficult to resolve small regions such as subregions of auditory cortex when anatomical morphology varies among subjects. Here, we use data from an auditory fMRI study to show that using a high-dimensional registration technique (HAMMER) results in an enhanced functional signal-to-noise ratio (fSNR) for functional data analysis within auditory regions, with more localized activation patterns. The technique is validated against DARTEL, a high-dimensional diffeomorphic registration, as well as against commonly used low-dimensional normalization techniques such as the techniques provided with SPM2 (cosine basis functions) and SPM5 (unified segmentation) software packages. We also systematically examine how spatial resolution of the template image and spatial smoothing of the functional data affect the results. Only the high-dimensional technique (HAMMER) appears to be able to capitalize on the excellent anatomical resolution of a single-subject reference template, and, as expected, smoothing increased fSNR, but at the cost of spatial resolution. In general, results demonstrate significant improvement in fSNR using HAMMER compared to analysis after normalization using DARTEL, or conventional normalization such as cosine basis function and unified segmentation in SPM, with more precisely localized activation foci, at least for activation in the region of auditory cortex.     

Artificial shifting of fMRI activation localized by volume- and surface-based analyses

Spatial smoothing is an important post-processing procedure that is used to increase the signal-to-noise ratio (SNR) of blood oxygenation level-dependent signals (BOLD) in common functional magnetic resonance imaging (fMRI) applications. However, recent studies have shown that smoothing artificially shifts probabilistic local maxima of fMRI activations. In this study, we show shifting of the localization of functional centers in hand motor areas of the cerebral cortex by three-dimensional isotropic Gaussian kernel smoothing or two-dimensional heat kernel smoothing in volume- and surface-based fMRI analyses. Activation maps derived from smoothed echo planar imaging (EPI) data by volume- and surface-based analyses were assigned to the nodes of individual cortical surface models, and local maxima in the primary motor area (M1) and the primary somatosensory cortex (S1) were compared with those derived from non-smoothed risk map analysis, which is commonly used in presurgical applications. For each analysis, the Euclidean and geodesic distances between the correlation coefficients of local maxima derived from smoothed and non-smoothed EPI data were measured. The results show that the correlation coefficients derived from the volume- and surface-based analyses were about 29.4% and 42.9% higher for smoothed than for non-smoothed risk map analyses, and show minimum shifting of localizations by 12.1 mm and 6.9 mm on average in Euclidean distance, respectively, and about 9.5 mm and 5.7 mm on average in geodesic distance, respectively.     

Automated brainstem co-registration (ABC) for MRI

Group data analysis in brainstem neuroimaging is predicated on accurate co-registration of anatomy. As the brainstem is comprised of many functionally heterogeneous nuclei densely situated adjacent to one another, relatively small errors in co-registration can manifest in increased variance or decreased sensitivity (or significance) in detecting activations. We have devised a 2-stage automated, reference mask guided registration technique (Automated Brainstem Co-registration, or ABC) for improved brainstem co-registration. Our approach utilized a brainstem mask dataset to weight an automated co-registration cost function. Our method was validated through measurement of RMS error at 12 manually defined landmarks. These landmarks were also used as guides for a secondary manual co-registration option, intended for outlier individuals that may not adequately co-register with our automated method. Our methodology was tested on 10 healthy human subjects and compared to traditional co-registration techniques (Talairach transform and automated affine transform to the MNI-152 template). We found that ABC had a significantly lower mean RMS error (1.22 +/- 0.39 mm) than Talairach transform (2.88 +/- 1.22 mm, mu +/- sigma) and the global affine (3.26 +/- 0.81 mm) method. Improved accuracy was also found for our manual-landmark-guided option (1.51 +/- 0.43 mm). Visualizing individual brainstem borders demonstrated more consistent and uniform overlap for ABC compared to traditional global co-registration techniques. Improved robustness (lower susceptibility to outliers) was demonstrated with ABC through lower inter-subject RMS error variance compared with traditional co-registration methods. The use of easily available and validated tools (AFNI and FSL) for this method should ease adoption by other investigators interested in brainstem data group analysis.     

Evaluation and comparison of GLM- and CVA-based fMRI processing pipelines with Java-based fMRI processing pipeline evaluation system

Activation patterns identified by fMRI processing pipelines or fMRI software packages are usually determined by the preprocessing options, parameters, and statistical models used. Previous studies that evaluated options of GLM (general linear model)--based fMRI processing pipelines are mainly based on simulated data with receiver operating characteristics (ROC) analysis, but evaluation of such fMRI processing pipelines on real fMRI data is rare. To understand the effect of processing options on performance of GLM-based fMRI processing pipelines with real fMRI data, we investigated the impact of commonly-used fMRI preprocessing steps; optimized the associated GLM-based single-subject processing pipelines; and quantitatively compared univariate GLM (in FSL.FEAT and NPAIRS.GLM) and multivariate CVA (canonical variates analysis) (in NPAIRS.CVA)-based analytic models in single-subject analysis with a recently developed fMRI processing pipeline evaluation system based on prediction accuracy (classification accuracy) and reproducibility performance metrics. For block-design data, we found that with GLM analysis (1) slice timing correction and global intensity normalization have little consistent impact on fMRI processing pipelines, spatial smoothing and high-pass filtering or temporal detrending significantly increases pipeline performance and thus are essential for robust fMRI statistical analysis; (2) combined optimization of spatial smoothing and temporal detrending improves pipeline performance; and (3) in general, the prediction performance of multivariate CVA is higher than that of the univariate GLM, while univariate GLM is more reproducible than multivariate CVA. Because of the different bias-variance trade-offs of univariate and multivariate models, it may be necessary to consider a consensus approach to obtain more accurate activation patterns in fMRI data.     

Evaluation of mixed effects in event-related fMRI studies: impact of first-level design and filtering

With the introduction of event-related designs in fMRI, it has become crucial to optimize design efficiency and temporal filtering to detect activations at the 1st level with high sensitivity. We investigate the relevance of these issues for fMRI population studies, that is, 2nd-level analysis, for a set of event-related fMRI (er-fMRI) designs with different 1st-level efficiencies, adopting three distinct 1st-level filtering strategies as implemented in SPM99, SPM2, and FSL3.0. By theory, experiments, and simulations using physiological fMRI noise, we show that both design and filtering impact the outcome of the statistical analysis, not only at the 1st but also at the 2nd level. There are several reasons behind this finding. First, sensitivity is affected by both design and filtering, since the scan-to-scan variance, that is the fixed effect, is not negligible with respect to the between-subject variance, that is the random effect, in er-fMRI population studies. The impact of the fixed effects error on the sensitivity of the mixed effects analysis can be mitigated by an optimal choice of er-fMRI design and filtering. Moreover, the accuracy of the 1st- and 2nd-level parameter estimates also depend on design and filtering; especially, we show that inaccuracies caused by the presence of residual noise autocorrelations can be constrained by designs that have hemodynamic responses with a Gaussian distribution. In conclusion, designs with both good efficiency and decorrelating properties, for example, such as the geometric or Latin square probability distributions, combined with the "whitening" filters of SPM2 and FSL3.0, give the best result, both for 1st- and 2nd-level analysis of er-fMRI studies.     

An empirical comparison of SPM preprocessing parameters to the analysis of fMRI data

We present the results from two sets of Monte Carlo simulations aimed at evaluating the robustness of some preprocessing parameters of SPM99 for the analysis of functional magnetic resonance imaging (fMRI). Statistical robustness was estimated by implementing parametric and nonparametric simulation approaches based on the images obtained from an event-related fMRI experiment. Simulated datasets were tested for combinations of the following parameters: basis function, global scaling, low-pass filter, high-pass filter and autoregressive modeling of serial autocorrelation. Based on single-subject SPM analysis, we derived the following conclusions that may serve as a guide for initial analysis of fMRI data using SPM99: (1) The canonical hemodynamic response function is a more reliable basis function to model the fMRI time series than HRF with time derivative. (2) Global scaling should be avoided since it may significantly decrease the power depending on the experimental design. (3) The use of a high-pass filter may be beneficial for event-related designs with fixed interstimulus intervals. (4) When dealing with fMRI time series with short interstimulus intervals (<8 s), the use of first-order autoregressive model is recommended over a low-pass filter (HRF) because it reduces the risk of inferential bias while providing a relatively good power. For datasets with interstimulus intervals longer than 8 seconds, temporal smoothing is not recommended since it decreases power. While the generalizability of our results may be limited, the methods we employed can be easily implemented by other scientists to determine the best parameter combination to analyze their data.     

Functional neuroimaging of the oculomotor brainstem network in humans

The cortical systems involved in eye movement control in humans have been investigated extensively using fMRI. In contrast, there is virtually no data concerning the functional status of the human oculomotor brainstem nuclei. This lack of evidence has usually been explained by technical constraints of EPI based imaging and anatomical characteristics of the brainstem. Against this assumption, we successfully localised nuclei of the oculomotor system using high-resolution fMRI based on standard EPI sequences in a group of healthy subjects executing reflexive horizontal saccades. A random-effects group analysis revealed task-related BOLD increases in the superior colliculus, the oculomotor nucleus, the abducens nucleus and in the paramedian pontine reticular formation. This group analysis was complemented by individual positive findings in up to 94% of single subject analyses. A visual control paradigm led to increased signal levels in the superior colliculus consistent with its visual properties but no corresponding signal changes in other brainstem nuclei. These results are consistent with findings in animal studies and demonstrate the feasibility to detect BOLD signal increases associated with oculomotor tasks even in the human brainstem using conventional EPI imaging techniques.     

Unified SPM-ICA for fMRI analysis

A widely used tool for functional magnetic resonance imaging (fMRI) data analysis, statistical parametric mapping (SPM), is based on the general linear model (GLM). SPM therefore requires a priori knowledge or specific assumptions about the time courses contributing to signal changes. In contradistinction, independent component analysis (ICA) is a data-driven method based on the assumption that the causes of responses are statistically independent. Here we describe a unified method, which combines ICA, temporal ICA (tICA), and SPM for analyzing fMRI data. tICA was applied to fMRI datasets to disclose independent components, whose number was determined by the Bayesian information criterion (BIC). The resulting components were used to construct the design matrix of a GLM. Parameters were estimated and regionally-specific statistical inferences were made about activations in the usual way. The sensitivity and specificity were evaluated using Monte Carlo simulations. The receiver operating characteristic (ROC) curves indicated that the unified SPM-ICA method had a better performance. Moreover, SPM-ICA was applied to fMRI datasets from twelve normal subjects performing left and right hand movements. The areas identified corresponded to motor (premotor, sensorimotor areas and SMA) areas and were consistently task related. Part of the frontal lobe, parietal cortex, and cingulate gyrus also showed transiently task-related responses. The unified method requires less supervision than the conventional SPM and enables classical inference about the expression of independent components. Our results also suggest that the method has a higher sensitivity than SPM analyses.     

Statistical parametric mapping: assessment of application in children

SPM is a powerful technique for the comparison of functional imaging data sets among groups of patients. While this technique has been widely applied in studies of adults, it has rarely been applied to studies of children, due in part to the lack of validation of the spatial normalization procedure in children of different ages. In order to determine if spatial normalization of FDG PET images using SPM96 to an adult template can be successfully applied in children, we applied PET-derived transformation parameters to coregistered MRI images. We then compared contours of spatially normalized MRI images obtained from 13 children with epilepsy (ages 2-14 years, mean 7.6 +/- 3.9 years) with those derived from 17 adult controls (mean age 27.6 +/- 4.5 years). Contours of spatially normalized MRI image volumes derived from the pediatric group were more variable than those obtained from adult controls. The average deviation from the mean adult contour was age-dependent and decreased with age (average deviation (mm) = 2.22 (mm) - 0.021 (mm/year) x years, r = 0.70, P < 0.001). Separate SPM analyses were performed for children less than 6 years (N1 = 6) and for children between 6 and 14 years of age (N2 = 7). SPM analyses performed in both pediatric groups showed significant regions of hypometabolism in locations consistent with their epileptic foci. SPM analyses in the younger group also showed significant artifacts. Therefore, the error associated with spatial normalization of pediatric brains to an adult template in children less than 6 years of age precludes the application of statistical parametric mapping in this age group. Although the error in the spatial normalization procedure for children ages 6 to 14 years is higher than in adults, it appears that this error does not result in artifacts in the SPM analysis. Furthermore, in contrast our previous studies showing large age-related changes in the absolute glucose metabolic rate at puberty, the SPM analysis showed children over 6 years of age appear to display the same pattern of glucose utilization as adults. However, small differences in the pattern of glucose utilization which might occur during late childhood and adolescence may not have been detected due to the sample size.     

Multi-site voxel-based morphometry--not quite there yet

Voxel-based morphometry (VBM) is a widely applied method in computational neurosciences but it is currently recommended to compare only data collected at a single MRI scanner. Multi-site VBM would be a desirable approach to increase group size and, thus, statistical power. We aimed to assess if multi-site VBM is feasible on similar hardware and compare the magnitude of inter- and intra-scanner differences. 18 healthy subjects were scanned in two identical 3T MRI scanners using different head coil designs, twice in scanner A and once in scanner B. 3D T1-weighted images were processed with SPM8 and FSL4.1 and compared as paired t-test (scan versus re-scan) on a voxel basis by means of a general linear model (GLM). Additionally, coefficient-of-difference (coeffD) maps were calculated for respective pairs of gray matter segmentations. We found considerable inter-scanner differences clearly exceeding a commonly used GLM significance threshold of p<0.05 (FWE corrected). The spatial pattern of detected differences was dependent on whether SPM8 or FSL4.1 was used. The inclusion of global correcting factors either aggravated (SPM8) or reduced the GLM detected differences (FSL4.1). The coeffD analysis revealed markedly higher variability within the FSL4.1 stream both for the inter- and the intra-scanner comparison. A lowered bias cutoff (30 mm FWHM) in SPM8 improved the comparability for cortical areas. Intra-scanner scan/re-scan differences were generally weaker and did not exceed a p<0.05 (FWE corrected) threshold in the GLM analysis. At 3T profound inter-scanner differences are to be expected that could severely confound an unbalanced VBM analysis. These are like related to the receive bias of the radio-frequency hardware.     

Comparison of fMRI motion correction software tools

Motion correction of fMRI data is a widely used step prior to data analysis. In this study, a comparison of the motion correction tools provided by several leading fMRI analysis software packages was performed, including AFNI, AIR, BrainVoyager, FSL, and SPM2. Comparisons were performed using data from typical human studies as well as phantom data. The identical reconstruction, preprocessing, and analysis steps were used on every data set, except that motion correction was performed using various configurations from each software package. Each package was studied using default parameters, as well as parameters optimized for speed and accuracy. Forty subjects performed a Go/No-go task (an event-related design that investigates inhibitory motor response) and an N-back task (a block-design paradigm investigating working memory). The human data were analyzed by extracting a set of general linear model (GLM)-derived activation results and comparing the effect of motion correction on thresholded activation cluster size and maximum t value. In addition, a series of simulated phantom data sets were created with known activation locations, magnitudes, and realistic motion. Results from the phantom data indicate that AFNI and SPM2 yield the most accurate motion estimation parameters, while AFNI's interpolation algorithm introduces the least smoothing. AFNI is also the fastest of the packages tested. However, these advantages did not produce noticeably better activation results in motion-corrected data from typical human fMRI experiments. Although differences in performance between packages were apparent in the human data, no single software package produced dramatically better results than the others. The "accurate" parameters showed virtually no improvement in cluster t values compared to the standard parameters. While the "fast" parameters did not result in a substantial increase in speed, they did not degrade the cluster results very much either. The phantom and human data indicate that motion correction can be a valuable step in the data processing chain, yielding improvements of up to 20% in the magnitude and up to 100% in the cluster size of detected activations, but the choice of software package does not substantially affect this improvement.     

Optimal slice timing correction and its interaction with fMRI parameters and artifacts

Due to the nature of fMRI acquisition protocols, slices in the plane of acquisition are not acquired simultaneously or sequentially, and therefore are temporally misaligned with each other. Slice timing correction (STC) is a critical preprocessing step that corrects for this temporal misalignment. Interpolation-based STC is implemented in all major fMRI processing software packages. To date, little effort has gone towards assessing the optimal method of STC. Delineating the benefits of STC can be challenging because of its slice-dependent gain as well as its interaction with other fMRI artifacts. In this study, we propose a new optimal method (Filter-Shift) based on the fundamental properties of sampling theory in digital signal processing. We then evaluate our method by comparing it to two other methods of STC from the most popular statistical software packages, SPM and FSL. STC methods were evaluated using 338 simulated and 30 real fMRI data and demonstrate the effectiveness of STC in general as well as the superiority of the proposed method in comparison to existing ones. All methods were evaluated under various scan conditions such as motion level, interleave sequence, scanner sampling rate, and the duration of the scan itself.     

采用独立分量分析观察经皮穴位电针刺激对脑功能的影响

目的使用独立分量分析方法探索督脉穴位经皮电刺激对脑功能的影响。方法使用1.5T GE Signa Excite核磁成像仪对一位女性脑外伤患者进行BOLD成像。采用组块设计,静息期与刺激期交替,组块长度均为30 s。数据处理采用GIFT、SPM5和MRIcro软件进行,并将独立分量分析与SPM软件处理的结果进行比较。结果采用GIFT中的扩展Infomax算法进行独立分量分析,显示有13个独立成分,每一独立成分包含一空间图和相应的时间变化曲线。任务相关性独立成分的空间激活图与SPM5的分析结果类似,但并不完全相同。此外,这些任务相关性独立成分的时间曲线与SPM所用的经典血流动力相应函数模型的形状并不一致。结论在使用模型依赖的数据分析方法如SPM之前,可以使用独立分量分析探索fMRI数据并获得先验知识。     

Enhancing reproducibility of fMRI statistical maps using generalized canonical correlation analysis in NPAIRS framework

Common fMRI data processing techniques usually minimize a temporal cost function or fit a temporal model to extract an activity map. Here, we focus on extracting a highly, spatially reproducible statistical parametric map (SPM) from fMRI data using a cost function that does not depend on a model of the subjects' temporal response. Based on a generalized version of canonical correlation analysis (gCCA), we propose a method to extract a highly reproducible map by maximizing the sum of pair-wise correlations between some maps. In a group analysis, each map is calculated from a linear combination of fMRI scans of a subset of subjects under study. The proposed method is applied to BOLD fMRI datasets without any spatial smoothing from 10 subjects performing a simple reaction time (RT) task. Using the NPAIRS split-half resampling framework with a reproducibility measure based on SPM correlations, we compare the proposed approach with canonical variate analysis (CVA) and a simple general linear model (GLM). gCCA provides statistical parametric maps with higher reproducibility than CVA and GLM with correlation reproducibilities across independent split-half SPMs of 0.78, 0.46, and 0.41, respectively. Our results show that gCCA is an efficient approach for extracting the default mode network, assessing brain connectivity, and processing event-related and resting-state datasets in which the temporal BOLD signal varies from subject to subject.     

The effect of slice order and thickness on fMRI activation data using multislice echo-planar imaging

Multislice echo-planar imaging (EPI) is a commonly used technique for fMRI studies. Brain activation images acquired using fMRI are sensitive to T2* changes, reflecting the level of blood oxygenation (BOLD contrast), and may also contain an element of T1 contrast which detects blood flow changes in large vessels. If slice inflow (T1) effects are significant in multislice EPI, then as the order in which the slices are acquired is changed, differences in the activation maps are predicted. However, in experiments presented here using visual stimulation, the data demonstrate that highly consistent results can be achieved for repetition times (TR) of 6.0, 3.0, and 1.5 s. This suggests that, for whole-brain multislice EPI, fMRI activation is dominated by T2*, BOLD contrast. The thickness of the imaging slice is also an important parameter in these studies, having implications for spatial resolution, sensitivity, and acquisition time. In separate visual cortex experiments the effect on the values of the fMRI Z scores and the number of activated voxels is investigated as a function of slice thickness (from 1 to 8 mm). The maximum Z scores in the data are similar for all slice thicknesses and, after resampling to allow a direct comparison to be made, the volume of visual cortex detected as significantly activated increases with slice thickness.     

Accuracy and sensitivity of detection of activation foci in the brain via statistical parametric mapping: a study using a PET simulator

Statistical parametric mapping (SPM) is currently the most widely used method for analysis of functional activation images. This paper reports a quantitative evaluation of the sensitivity and accuracy of SPM, using a realistic simulator of PET image formation, which accounted for the main physical processes involved in PET, including attenuation, scatter, randoms, Poisson noise, and limited detector resolution. Activation foci of the brain were simulated by placing spheres of specified activities in particular locations. Using these data, the sensitivity and accuracy of SPM in detecting activation foci was measured for different versions of the SPM spatial normalization method and for an elastic warping method referred to as STAR (spatial transformation algorithm for registration). The STAR method resulted in relatively better registration and hence better detection of the activation foci. A secondary goal of the paper was to evaluate the improvement in detection sensitivity obtained by applying an atlas-based adaptive smoothing method instead of the usual Gaussian filtering method. The results indicate some limitations of statistical parametric mapping, assist in the correct interpretation of the SPM maps, and point to future research directions in functional image analysis.     

Reduction of gradient acoustic noise in MRI using SENSE-EPI

A new approach to reduce gradient acoustic noise levels in EPI experiments is presented. Using multichannel RF receive coils, combined with SENSE data acquisition and reconstruction, gradient slew-rates in single-shot EPI were reduced fourfold for rate-2 and ninefold for rate-3 SENSE. Multislice EPI experiments were performed on three different scanner platforms. With 3.4 mm in-plane resolution, measuring 6 slices per second (12 slices with 2000 ms TR), this resulted in average sound pressure level reductions of 11.3 dB(A) and 16.5 dB(A) for rate-2 and rate-3 SENSE, respectively. BOLD fMRI experiments, using visually paced finger-tapping paradigms, showed no detrimental effect of the acoustic noise reduction strategy on temporal noise levels and t scores.     

基于EPI方法的功能磁共振成像质量问题实例分析:主要成因与应对方案

基于回波平面成像(echo planar image,EPI)序列的功能磁共振成像(functional magnetic resonance imaging,fMRI)是当前基础与临床脑功能研究实现快速成像的主要手段。但是,与常规成像序列相比,这种快速成像更容易受到各种噪声和伪影的干扰,从而影响fMRI的数据质量。如何及时发现和解决EPI的伪影和稳定性问题,是fMRI科研数据质量控制的一个核心问题。本文对笔者在过去五年fMRI质量控制工作中发现的成像问题进行整理,总结了fMRI应用中经常遇到的五种主要伪影,包括奈奎斯特鬼影、几何畸变、尖峰噪声、射频噪声和EPI稳定性相关的问题。并为每种伪影选取了具有代表性的案例,分析它们的特点、主要成因和应对方案。希望这些案例和分析结果能为fMRI用户和技术人员提供有益的借鉴。