A survey of the sources of noise in fMRI

Precision fMRI and cluster-failure in the individual brain

Advances in neuroimaging acquisition protocols and denoising techniques, along with increasing magnetic field strengths, have dramatically improved the temporal signal-to-noise ratio (tSNR) in functional magnetic resonance imaging (fMRI). This permits spatial resolution with submillimeter voxel sizes and ultrahigh temporal resolution and opens a route toward performing precision fMRI in the brains of individuals. Yet ultrahigh spatial and temporal resolution comes at a cost: it reduces tSNR and, therefore, the sensitivity to the blood oxygen level-dependent (BOLD) effect and other functional contrasts across the brain. Here we investigate the potential of various smoothing filters to improve BOLD sensitivity while preserving the spatial accuracy of activated clusters in single-subject analysis. We introduce adaptive-weight smoothing with optimized metrics (AWSOM), which addresses this challenge extremely well. AWSOM employs a local inference approach that is as sensitive as cluster-corrected inference of data smoothed with large Gaussian kernels, but it preserves spatial details across multiple tSNR levels. This is essential for examining whole-brain fMRI data because tSNR varies across the entire brain, depending on the distance of a brain region from the receiver coil, the type of setup, acquisition protocol, preprocessing, and resolution. We found that cluster correction in single subjects results in inflated family-wise error and false positive rates. AWSOM effectively suppresses false positives while remaining sensitive even to small clusters of activated voxels. Furthermore, it preserves signal integrity, that is, the relative activation strength of significant voxels, making it a valuable asset for a wide range of fMRI applications. Here we demonstrate these features and make AWSOM freely available to the research community for download.

Connectome predictive modeling of trait mindfulness

Introduction

Trait mindfulness refers to one's disposition or tendency to pay attention to their experiences in the present moment, in a non-judgmental and accepting way. Trait mindfulness has been robustly associated with positive mental health outcomes, but its neural underpinnings are poorly understood. Prior resting-state fMRI studies have associated trait mindfulness with within- and between-network connectivity of the default-mode (DMN), fronto-parietal (FPN), and salience networks. However, it is unclear how generalizable the findings are, how they relate to different components of trait mindfulness, and how other networks and brain areas may be involved.

Methods

To address these gaps, we conducted the largest resting-state fMRI study of trait mindfulness to-date, consisting of a pre-registered connectome predictive modeling analysis in 367 adults across three samples collected at different sites.

Results

In the model-training dataset, we did not find connections that predicted overall trait mindfulness, but we identified neural models of two mindfulness subscales, Acting with Awareness and Non-judging. Models included both positive networks (sets of pairwise connections that positively predicted mindfulness with increasing connectivity) and negative networks, which showed the inverse relationship. The Acting with Awareness and Non-judging positive network models showed distinct network representations involving FPN and DMN, respectively. The negative network models, which overlapped significantly across subscales, involved connections across the whole brain with prominent involvement of somatomotor, visual and DMN networks. Only the negative networks generalized to predict subscale scores out-of-sample, and not across both test datasets. Predictions from both models were also negatively correlated with predictions from a well-established mind-wandering connectome model.

Conclusions

We present preliminary neural evidence for a generalizable connectivity models of trait mindfulness based on specific affective and cognitive facets. However, the incomplete generalization of the models across all sites and scanners, limited stability of the models, as well as the substantial overlap between the models, underscores the difficulty of finding robust brain markers of mindfulness facets.

The relative brain signal variability increases in the behavioral variant of frontotemporal dementia and Alzheimer's disease but not in schizophrenia

Overlapping symptoms between Alzheimer's disease (AD), behavioral variant of frontotemporal dementia (bvFTD), and schizophrenia (SZ) can lead to misdiagnosis and delays in appropriate treatment, especially in cases of early-onset dementia. To determine the potential of brain signal variability as a diagnostic tool, we assessed the coefficient of variation of the BOLD signal (CVBOLD) in 234 participants spanning bvFTD (n = 53), AD (n = 17), SZ (n = 23), and controls (n = 141). All underwent functional and structural MRI scans. Data unveiled a notable increase in CVBOLD in bvFTD patients across both datasets (local and international, p < 0.05), revealing an association with clinical scores (CDR and MMSE, r = 0.46 and r = -0.48, p < 0.0001). While SZ and control group demonstrated no significant differences, a comparative analysis between AD and bvFTD patients spotlighted elevated CVBOLD in the frontopolar cortices for the latter (p < 0.05). Furthermore, CVBOLD not only presented excellent diagnostic accuracy for bvFTD (AUC 0.78-0.95) but also showcased longitudinal repeatability. During a one-year follow-up, the CVBOLD levels increased by an average of 35% in the bvFTD group, compared to a 2% increase in the control group (p < 0.05). Our findings suggest that CVBOLD holds promise as a biomarker for bvFTD, offering potential for monitoring disease progression and differentiating bvFTD from AD and SZ.

Macroscopic resting-state brain dynamics are best described by linear models

It is typically assumed that large networks of neurons exhibit a large repertoire of nonlinear behaviours. Here we challenge this assumption by leveraging mathematical models derived from measurements of local field potentials via intracranial electroencephalography and of whole-brain blood-oxygen-level-dependent brain activity via functional magnetic resonance imaging. We used state-of-the-art linear and nonlinear families of models to describe spontaneous resting-state activity of 700 participants in the Human Connectome Project and 122 participants in the Restoring Active Memory project. We found that linear autoregressive models provide the best fit across both data types and three performance metrics: predictive power, computational complexity and the extent of the residual dynamics unexplained by the model. To explain this observation, we show that microscopic nonlinear dynamics can be counteracted or masked by four factors associated with macroscopic dynamics: averaging over space and over time, which are inherent to aggregated macroscopic brain activity, and observation noise and limited data samples, which stem from technological limitations. We therefore argue that easier-to-interpret linear models can faithfully describe macroscopic brain dynamics during resting-state conditions.

Denoising task-related fMRI: Balancing noise reduction against signal loss

Preprocessing fMRI data requires striking a fine balance between conserving signals of interest and removing noise. Typical steps of preprocessing include motion correction, slice timing correction, spatial smoothing, and high-pass filtering. However, these standard steps do not remove many sources of noise. Thus, noise-reduction techniques, for example, CompCor, FIX, and ICA-AROMA have been developed to further improve the ability to draw meaningful conclusions from the data. The ability of these techniques to minimize noise while conserving signals of interest has been tested almost exclusively in resting-state fMRI and, only rarely, in task-related fMRI. Application of noise-reduction techniques to task-related fMRI is particularly important given that such procedures have been shown to reduce false positive rates. Little remains known about the impact of these techniques on the retention of signal in tasks that may be associated with systemic physiological changes. In this paper, we compared two ICA-based, that is FIX and ICA-AROMA, two CompCor-based noise-reduction techniques, that is aCompCor, and tCompCor, and standard preprocessing using a large (n = 101) fMRI dataset including noxious heat and non-noxious auditory stimulation. Results show that preprocessing using FIX performs optimally for data obtained using noxious heat, conserving more signals than CompCor-based techniques and ICA-AROMA, while removing only slightly less noise. Similarly, for data obtained during non-noxious auditory stimulation, FIX noise-reduction technique before analysis with a covariate of interest outperforms the other techniques. These results indicate that FIX might be the most appropriate technique to achieve the balance between conserving signals of interest and removing noise during task-related fMRI.

Inter-regional correlation estimators for functional magnetic resonance imaging

Functional magnetic resonance imaging (fMRI) functional connectivity between brain regions is often computed using parcellations defined by functional or structural atlases. Typically, some kind of voxel averaging is performed to obtain a single temporal correlation estimate per region pair. However, several estimators can be defined for this task, with various assumptions and degrees of robustness to local noise, global noise, and region size. In this paper, we systematically present and study the properties of 9 different functional connectivity estimators taking into account the spatial structure of fMRI data, based on a simple fMRI data spatial model. These include 3 existing estimators and 6 novel estimators. We demonstrate the empirical properties of the estimators using synthetic, animal, and human data, in terms of graph structure, repeatability and reproducibility, discriminability, dependence on region size, as well as local and global noise robustness. We prove analytically the link between regional intra-correlation and inter-region correlation, and show that the choice of estimator has a strong influence on inter-correlation values. Some estimators, including the commonly used correlation of averages (ca), are positively biased, and have more dependence to region size and intra-correlation than robust alternatives, resulting in spatially-dependent bias. We define the new local correlation of averages estimator with better theoretical guarantees, lower bias, significantly lower dependence on region size (Spearman correlation 0.40 vs 0.55, paired t-test T=27.2, p=1.1e-47), at negligible cost to discriminative power, compared to the ca estimator. The difference in connectivity pattern between the estimators is not distributed uniformly throughout the brain, but rather shows a clear ventral-dorsal gradient, suggesting that region size and intra-correlation plays an important role in shaping functional networks defined using the ca estimator, and leading to non-trivial differences in their connectivity structure. We provide an open source R package and equivalent Python implementation to facilitate the use of the new estimators, together with preprocessed rat time-series.

Model connectivity: leveraging the power of encoding models to overcome the limitations of functional connectivity

Functional connectivity (FC) is the most popular method for recovering functional networks of brain areas with fMRI. However, because FC is defined as temporal correlations in brain activity, FC networks are confounded by noise and lack a precise functional role. To overcome these limitations, we developed model connectivity (MC). MC is defined as similarities in encoding model weights, which quantify reliable functional activity in terms of interpretable stimulus- or task-related features. To compare FC and MC, both methods were applied to a naturalistic story listening dataset. FC recovered spatially broad networks that are confounded by noise, and that lack a clear role during natural language comprehension. By contrast, MC recovered spatially localized networks that are robust to noise, and that represent distinct categories of semantic concepts. Thus, MC is a powerful data-driven approach for recovering and interpreting the functional networks that support complex cognitive processes.

Functional connectivity MRI quality control procedures in CONN

Quality control (QC) for functional connectivity magnetic resonance imaging (FC-MRI) is critical to ensure the validity of neuroimaging studies. Noise confounds are common in MRI data and, if not accounted for, may introduce biases in functional measures affecting the validity, replicability, and interpretation of FC-MRI study results. Although FC-MRI analysis rests on the assumption of adequate data processing, QC is underutilized and not systematically reported. Here, we describe a quality control pipeline for the visual and automated evaluation of MRI data implemented as part of the CONN toolbox. We analyzed publicly available resting state MRI data (N = 139 from 7 MRI sites) from the FMRI Open QC Project. Preprocessing steps included realignment, unwarp, normalization, segmentation, outlier identification, and smoothing. Data denoising was performed based on the combination of scrubbing, motion regression, and aCompCor - a principal component characterization of noise from minimally eroded masks of white matter and of cerebrospinal fluid tissues. Participant-level QC procedures included visual inspection of raw-level data and of representative images after each preprocessing step for each run, as well as the computation of automated descriptive QC measures such as average framewise displacement, average global signal change, prevalence of outlier scans, MNI to anatomical and functional overlap, anatomical to functional overlap, residual BOLD timeseries variability, effective degrees of freedom, and global correlation strength. Dataset-level QC procedures included the evaluation of inter-subject variability in the distributions of edge connectivity in a 1,000-node graph (FC distribution displays), and the estimation of residual associations across participants between functional connectivity strength and potential noise indicators such as participant's head motion and prevalence of outlier scans (QC-FC analyses). QC procedures are demonstrated on the reference dataset with an emphasis on visualization, and general recommendations for best practices are discussed in the context of functional connectivity and other fMRI analysis. We hope this work contributes toward the dissemination and standardization of QC testing performance reporting among peers and in scientific journals.

Multi-Echo Investigations of Positive and Negative CBF and Concomitant BOLD Changes: Positive and negative CBF and BOLD changes

Unlike the positive blood oxygenation level-dependent (BOLD) response (PBR), commonly taken as an indication of an 'activated' brain region, the physiological origin of negative BOLD signal changes (i.e. a negative BOLD response, NBR), also referred to as 'deactivation' is still being debated. In this work, an attempt was made to gain a better understanding of the underlying mechanism by obtaining a comprehensive measure of the contributing cerebral blood flow (CBF) and its relationship to the NBR in the human visual cortex, in comparison to a simultaneously induced PBR in surrounding visual regions. To overcome the low signal-to-noise ratio (SNR) of CBF measurements, a newly developed multi-echo version of a center-out echo planar-imaging (EPI) readout was employed with pseudo-continuous arterial spin labeling (pCASL). It achieved very short echo and inter-echo times and facilitated a simultaneous detection of functional CBF and BOLD changes at 3 T with improved sensitivity. Evaluations of the absolute and relative changes of CBF and the effective transverse relaxation rate,R₂^* the coupling ratios, and their dependence on CBF at rest, CBF_rest indicated differences between activated and deactivated regions. Analysis of the shape of the respective functional responses also revealed faster negative responses with more pronounced post-stimulus transients. Resulting differences in the flow-metabolism coupling ratios were further examined for potential distinctions in the underlying neuronal contributions.

A hands-on tutorial on network and topological neuroscience

The brain is an extraordinarily complex system that facilitates the optimal integration of information from different regions to execute its functions. With the recent advances in technology, researchers can now collect enormous amounts of data from the brain using neuroimaging at different scales and from numerous modalities. With that comes the need for sophisticated tools for analysis. The field of network neuroscience has been trying to tackle these challenges, and graph theory has been one of its essential branches through the investigation of brain networks. Recently, topological data analysis has gained more attention as an alternative framework by providing a set of metrics that go beyond pairwise connections and offer improved robustness against noise. In this hands-on tutorial, our goal is to provide the computational tools to explore neuroimaging data using these frameworks and to facilitate their accessibility, data visualisation, and comprehension for newcomers to the field. We will start by giving a concise (and by no means complete) overview of the field to introduce the two frameworks and then explain how to compute both well-established and newer metrics on resting-state functional magnetic resonance imaging. We use an open-source language (Python) and provide an accompanying publicly available Jupyter Notebook that uses the 1000 Functional Connectomes Project dataset. Moreover, we would like to highlight one part of our notebook dedicated to the realistic visualisation of high order interactions in brain networks. This pipeline provides three-dimensional (3-D) plots of pairwise and higher-order interactions projected in a brain atlas, a new feature tailor-made for network neuroscience.

Imaging faster neural dynamics with fast fMRI: A need for updated models of the hemodynamic response

Fast fMRI enables the detection of neural dynamics over timescales of hundreds of milliseconds, suggesting it may provide a new avenue for studying subsecond neural processes in the human brain. The magnitudes of these fast fMRI dynamics are far greater than predicted by canonical models of the hemodynamic response. Several studies have established nonlinear properties of the hemodynamic response that have significant implications for fast fMRI. We first review nonlinear properties of the hemodynamic response function that may underlie fast fMRI signals. We then illustrate the breakdown of canonical hemodynamic response models in the context of fast neural dynamics. We will then argue that the canonical hemodynamic response function is not likely to reflect the BOLD response to neuronal activity driven by sparse or naturalistic stimuli or perhaps to spontaneous neuronal fluctuations in the resting state. These properties suggest that fast fMRI is capable of tracking surprisingly fast neuronal dynamics, and we discuss the neuroscientific questions that could be addressed using this approach.

A practical modification to a resting state fMRI protocol for improved characterization of cerebrovascular function

Cerebrovascular reactivity (CVR), defined here as the Blood Oxygenation Level Dependent (BOLD) response to a CO₂ pressure change, is a useful metric of cerebrovascular function. Both the amplitude and the timing (hemodynamic lag) of the CVR response can bring insight into the nature of a cerebrovascular pathology and aid in understanding noise confounds when using functional Magnetic Resonance Imaging (fMRI) to study neural activity. This research assessed a practical modification to a typical resting-state fMRI protocol, to improve the characterization of cerebrovascular function. In 9 healthy subjects, we modelled CVR and lag in three resting-state data segments, and in data segments which added a 2–3 minute breathing task to the start of a resting-state segment. Two different breathing tasks were used to induce fluctuations in arterial CO₂ pressure: a breath-hold task to induce hypercapnia (CO₂ increase) and a cued deep breathing task to induce hypocapnia (CO₂ decrease). Our analysis produced voxel-wise estimates of the amplitude (CVR) and timing (lag) of the BOLD-fMRI response to CO₂ by systematically shifting the CO₂ regressor in time to optimize the model fit. This optimization inherently increases gray matter CVR values and fit statistics. The inclusion of a simple breathing task, compared to a resting-state scan only, increases the number of voxels in the brain that have a significant relationship between CO₂ and BOLD-fMRI signals, and improves our confidence in the plausibility of voxel-wise CVR and hemodynamic lag estimates. We demonstrate the clinical utility and feasibility of this protocol in an incidental finding of Moyamoya disease, and explore the possibilities and challenges of using this protocol in younger populations. This hybrid protocol has direct applications for CVR mapping in both research and clinical settings and wider applications for fMRI denoising and interpretation.

Ground-truth "resting-state" signal provides data-driven estimation and correction for scanner distortion of fMRI time-series dynamics

The fMRI community has made great strides in decoupling neuronal activity from other physiologically induced T₂* changes, using sensors that provide a ground-truth with respect to cardiac, respiratory, and head movement dynamics. However, blood oxygenation level-dependent (BOLD) time-series dynamics are also confounded by scanner artifacts, in complex ways that can vary not only between scanners but even, for the same scanner, between sessions. Unfortunately, the lack of an equivalent ground truth for BOLD time-series has thus far stymied the development of reliable methods for identification and removal of scanner-induced noise, a problem that we have previously shown to severely impact detection sensitivity of resting-state brain networks. To address this problem, we first designed and built a phantom capable of providing dynamic signals equivalent to that of the resting-state brain. Using the dynamic phantom, we then compared the ground-truth time-series with its measured fMRI data. Using these, we introduce data-quality metrics: Standardized Signal-to-Noise Ratio (ST-SNR) and Dynamic Fidelity that, unlike currently used measures such as temporal SNR (tSNR), can be directly compared across scanners. Dynamic phantom data acquired from four "best-case" scenarios: high-performance scanners with MR-physicist-optimized acquisition protocols, still showed scanner instability/multiplicative noise contributions of about 6-18% of the total noise. We further measured strong non-linearity in the fMRI response for all scanners, ranging between 8-19% of total voxels. To correct scanner distortion of fMRI time-series dynamics at a single-subject level, we trained a convolutional neural network (CNN) on paired sets of measured vs. ground-truth data. The CNN learned the unique features of each session's noise, providing a customized temporal filter. Tests on dynamic phantom time-series showed a 4- to 7-fold increase in ST-SNR and about 40-70% increase in Dynamic Fidelity after denoising, with CNN denoising outperforming both the temporal bandpass filtering and denoising using Marchenko-Pastur principal component analysis. Critically, we observed that the CNN temporal denoising pushes ST-SNR to a regime where signal power is higher than that of noise (ST-SNR > 1). Denoising human-data with ground-truth-trained CNN, in turn, showed markedly increased detection sensitivity of resting-state networks. These were visible even at the level of the single-subject, as required for clinical applications of fMRI.

Hemodynamic response varies across tactile stimuli with different temporal structures

Tactile stimuli can be distinguished based on their temporal features (e.g., duration, local frequency, and number of pulses), which are fundamental for vibrotactile frequency perception. Characterizing how the hemodynamic response changes in shape across experimental conditions is important for designing and interpreting fMRI studies on tactile information processing. In this study, we focused on periodic tactile stimuli with different temporal structures and explored the hemodynamic response function (HRF) induced by these stimuli. We found that HRFs were stimulus‐dependent in tactile‐related brain areas. Continuous stimuli induced a greater area of activation and a stronger and narrower hemodynamic response than intermittent stimuli with the same duration. The magnitude of the HRF increased with increasing stimulus duration. By normalizing the characteristics into topographic matrix, nonlinearity was obvious. These results suggested that stimulation patterns and duration within a cycle may be key characters for distinguishing different stimuli. We conclude that different temporal structures of tactile stimuli induced different HRFs, which are essential for vibrotactile perception and should be considered in fMRI experimental designs and analyses.

Estimating differential latent variable graphical models with applications to brain connectivity

Differential graphical models are designed to represent the difference between the conditional dependence structures of two groups, and thus are of particular interest for scientific investigations. Motivated by modern applications, this manuscript considers an extended setting where each group is generated by a latent variable Gaussian graphical model. Due to the existence of latent factors, the differential network is decomposed into sparse and low-rank components, both of which are symmetric indefinite matrices. We estimate these two components simultaneously using a two-stage procedure: (i) an initialization stage, which computes a simple, consistent estimator, and (ii) a convergence stage, implemented using a projected alternating gradient descent algorithm applied to a nonconvex objective, initialized using the output of the first stage. We prove that given the initialization, the estimator converges linearly with a nontrivial, minimax optimal statistical error. Experiments on synthetic and real data illustrate that the proposed nonconvex procedure outperforms existing methods.

Confound modelling in UK Biobank brain imaging

Dealing with confounds is an essential step in large cohort studies to address problems such as unexplained variance and spurious correlations. UK Biobank is a powerful resource for studying associations between imaging and non-imaging measures such as lifestyle factors and health outcomes, in part because of the large subject numbers. However, the resulting high statistical power also raises the sensitivity to confound effects, which therefore have to be carefully considered. In this work we describe a set of possible confounds (including nonlinear effects and interactions that researchers may wish to consider for their studies using such data). We include descriptions of how we can estimate the confounds, and study the extent to which each of these confounds affects the data, and the spurious correlations that may arise if they are not controlled. Finally, we discuss several issues that future studies should consider when dealing with confounds.

Vigilance Effects in Resting-State fMRI

Measures of resting-state functional magnetic resonance imaging (rsfMRI) activity have been shown to be sensitive to cognitive function and disease state. However, there is growing evidence that variations in vigilance can lead to pronounced and spatially widespread differences in resting-state brain activity. Unless properly accounted for, differences in vigilance can give rise to changes in resting-state activity that can be misinterpreted as primary cognitive or disease-related effects. In this paper, we examine in detail the link between vigilance and rsfMRI measures, such as signal variance and functional connectivity. We consider how state changes due to factors such as caffeine and sleep deprivation affect both vigilance and rsfMRI measures and review emerging approaches and methodological challenges for the estimation and interpretation of vigilance effects.

The Roles of Statistics in Human Neuroscience

Statistics plays three important roles in brain studies. They are (1) the study of differences between brains in distinctive populations; (2) the study of the variability in the structure and functioning of the brain; and (3) the study of data reduction on large-scale brain data. I discuss these concepts using examples from past and ongoing research in brain connectivity, brain information flow, information extraction from large-scale neuroimaging data, and neural predictive modeling. Having dispensed with the past, I attempt to present a few areas where statistical science facilitates brain decoding and to write prospectively, in the light of present knowledge and in the quest for artificial intelligence, about questions that statistical and neurobiological communities could work closely together to address in the future.

Cluster failure revisited: Impact of first level design and physiological noise on cluster false positive rates

Methodological research rarely generates a broad interest, yet our work on the validity of cluster inference methods for functional magnetic resonance imaging (fMRI) created intense discussion on both the minutia of our approach and its implications for the discipline. In the present work, we take on various critiques of our work and further explore the limitations of our original work. We address issues about the particular event-related designs we used, considering multiple event types and randomization of events between subjects. We consider the lack of validity found with one-sample permutation (sign flipping) tests, investigating a number of approaches to improve the false positive control of this widely used procedure. We found that the combination of a two-sided test and cleaning the data using ICA FIX resulted in nominal false positive rates for all data sets, meaning that data cleaning is not only important for resting state fMRI, but also for task fMRI. Finally, we discuss the implications of our work on the fMRI literature as a whole, estimating that at least 10% of the fMRI studies have used the most problematic cluster inference method (p = .01 cluster defining threshold), and how individual studies can be interpreted in light of our findings. These additional results underscore our original conclusions, on the importance of data sharing and thorough evaluation of statistical methods on realistic null data.

Stimulus-dependent hemodynamic response timing across the human subcortical-cortical visual pathway identified through high spatiotemporal resolution 7T fMRI

Recent developments in fMRI acquisition techniques now enable fast sampling with whole-brain coverage, suggesting fMRI can be used to track changes in neural activity at increasingly rapid timescales. When images are acquired at fast rates, the limiting factor for fMRI temporal resolution is the speed of the hemodynamic response. Given that HRFs may vary substantially in subcortical structures, characterizing the speed of subcortical hemodynamic responses, and how the hemodynamic response shape changes with stimulus duration (i.e. the hemodynamic nonlinearity), is needed for designing and interpreting fast fMRI studies of these regions. We studied the temporal properties and nonlinearities of the hemodynamic response function (HRF) across the human subcortical visual system, imaging superior colliculus (SC), lateral geniculate nucleus of the thalamus (LGN) and primary visual cortex (V1) with high spatiotemporal resolution 7 Tesla fMRI. By presenting stimuli of varying durations, we mapped the timing and nonlinearity of hemodynamic responses in these structures at high spatiotemporal resolution. We found that the hemodynamic response is consistently faster and narrower in subcortical structures than in cortex. However, the nonlinearity in LGN is similar to that in cortex, with shorter duration stimuli eliciting larger and faster responses than would have been predicted by a linear model. Using oscillatory visual stimuli, we tested the frequency response in LGN and found that its BOLD response tracked high-frequency (0.5 Hz) oscillations. The LGN response magnitudes were comparable to V1, allowing oscillatory BOLD signals to be detected in LGN despite the small size of this structure. These results suggest that the increase in the speed and amplitude of the hemodynamic response when neural activity is brief may be the key physiological driver of fast fMRI signals, enabling detection of high-frequency oscillations with fMRI. We conclude that subcortical visual structures exhibit fast and nonlinear hemodynamic responses, and that these dynamics enable detection of fast BOLD signals even within small deep brain structures when imaging is performed at ultra-high field.

Reprint of 'Noise contributions to the fMRI signal: An Overview'

The ability to discriminate signal from noise plays a key role in the analysis and interpretation of functional magnetic resonance imaging (fMRI) measures of brain activity. Over the past two decades, a number of major sources of noise have been identified, including system-related instabilities, subject motion, and physiological fluctuations. This article reviews the characteristics of the various noise sources as well as the mechanisms through which they affect the fMRI signal. Approaches for distinguishing signal from noise and the associated challenges are also reviewed. These challenges reflect the fact that some noise sources, such as respiratory activity, are generated by the same underlying brain networks that give rise to functional signals that are of interest.

Analysis strategies for high-resolution UHF-fMRI data

Functional MRI (fMRI) benefits from both increased sensitivity and specificity with increasing magnetic field strength, making it a key application for Ultra-High Field (UHF) MRI scanners. Most UHF-fMRI studies utilize the dramatic increases in sensitivity and specificity to acquire high-resolution data reaching sub-millimeter scales, which enable new classes of experiments to probe the functional organization of the human brain. This review article surveys advanced data analysis strategies developed for high-resolution fMRI at UHF. These include strategies designed to mitigate distortion and artifacts associated with higher fields in ways that attempt to preserve spatial resolution of the fMRI data, as well as recently introduced analysis techniques that are enabled by these extremely high-resolution data. Particular focus is placed on anatomically-informed analyses, including cortical surface-based analysis, which are powerful techniques that can guide each step of the analysis from preprocessing to statistical analysis to interpretation and visualization. New intracortical analysis techniques for laminar and columnar fMRI are also reviewed and discussed. Prospects for single-subject individualized analyses are also presented and discussed. Altogether, there are both specific challenges and opportunities presented by UHF-fMRI, and the use of proper analysis strategies can help these valuable data reach their full potential.

Using Functional Connectivity Magnetic Resonance Imaging to Measure Brain Connectivity in Preterm Infants

BACKGROUND

The use of functional connectivity magnetic resonance imaging (fcMRI) in research involving preterm infants is relatively new, and its feasibility in this population is not fully established. However, fcMRI images reveal functional neural connections that may be useful in establishing the mechanisms of neuroprotective interventions in preterm infants.

OBJECTIVE

The aim of this study was to determine the feasibility of using fcMRI to measure differences in functional neural connections in nursing intervention studies.

METHODS

A pilot study was conducted as part of a longitudinal, randomized controlled trial (RCT) testing the effect of a feeding intervention on neurodevelopmental and clinical outcomes of preterm infants randomly assigned to one of two groups: a patterned feeding experience (PFE) group and a usual feeding care (UFC) group. The fcMRIs were done at term-equivalent age. Visual, motor, and default mode networks were analyzed.

RESULTS

Seven infants were studied (four were in the PFE group, and three were in the UFC group). Participants were selected sequentially from the parent RCT. Clear images were obtained from all participants. Differences were noted among PFE and UFC infants: Infants receiving PFE were hyperconnective in the default mode (caudate, anterior cingulate cortex, and precuneus) and motor networks (middle temporal and middle occipital areas) and hypoconnective in others areas of the default mode (hippocampal and lingual regions) and motor networks (precentral and superior frontal cortices) relative to UFC infants. No differences were noted in visual networks.

DISCUSSION

The feasibility of using fcMRI at term-equivalent age in preterm infants who participated in an RCT on the effect of a nursing intervention was shown. Differences in connectivity among infants by group were detected. Further research is needed to show the benefit of fcMRI in studies of preterm infants given the costs of the procedure as well as the uncertain relationship of this early outcome measure to long-term neurodevelopment.

Modeling correlated noise is necessary to decode uncertainty

Brain decoding algorithms form an important part of the arsenal of analysis tools available to neuroscientists, allowing for a more detailed study of the kind of information represented in patterns of cortical activity. While most current decoding algorithms focus on estimating a single, most likely stimulus from the pattern of noisy fMRI responses, the presence of noise causes this estimate to be uncertain. This uncertainty in stimulus estimates is a potentially highly relevant aspect of cortical stimulus processing, and features prominently in Bayesian or probabilistic models of neural coding. Here, we focus on sensory uncertainty and how best to extract this information with fMRI. We first demonstrate in simulations that decoding algorithms that take into account correlated noise between fMRI voxels better recover the amount of uncertainty (quantified as the width of a probability distribution over possible stimuli) associated with the decoded estimate. Furthermore, we show that not all correlated variability should be treated equally, as modeling tuning-dependent correlations has the greatest impact on decoding performance. Next, we examine actual noise correlations in human visual cortex, and find that shared variability in areas V1-V3 depends on the tuning properties of fMRI voxels. In line with our simulations, accounting for this shared noise between similarly tuned voxels produces important benefits in decoding. Our findings underscore the importance of accurate noise models in fMRI decoding approaches, and suggest a statistically feasible method to incorporate the most relevant forms of shared noise.

Learning-based structurally-guided construction of resting-state functional correlation tensors

Functional magnetic resonance imaging (fMRI) measures changes in blood-oxygenation-level-dependent (BOLD) signals to detect brain activities. It has been recently reported that the spatial correlation patterns of resting-state BOLD signals in the white matter (WM) also give WM information often measured by diffusion tensor imaging (DTI). These correlation patterns can be captured using functional correlation tensor (FCT), which is analogous to the diffusion tensor (DT) obtained from DTI. In this paper, we propose a noise-robust FCT method aiming at further improving its quality, and making it eligible for further neuroscience study. The novel FCT estimation method consists of three major steps: First, we estimate the initial FCT using a patch-based approach for BOLD signal correlation to improve the noise robustness. Second, by utilizing the relationship between functional and diffusion data, we employ a regression forest model to learn the mapping between the initial FCTs and the corresponding DTs using the training data. The learned forest can then be applied to predict the DTI-like tensors given the initial FCTs from the testing fMRI data. Third, we re-estimate the enhanced FCT by utilizing the DTI-like tensors as a feedback guidance to further improve FCT computation. We have demonstrated the utility of our enhanced FCTs in Alzheimer's disease (AD) diagnosis by identifying mild cognitive impairment (MCI) patients from normal subjects.

Minimizing noise in pediatric task-based functional MRI; Adolescents with developmental disabilities and typical development

Functional Magnetic Resonance Imaging (fMRI) represents a powerful tool with which to examine brain functioning and development in typically developing pediatric groups as well as children and adolescents with clinical disorders. However, fMRI data can be highly susceptible to misinterpretation due to the effects of excessive levels of noise, often related to head motion. Imaging children, especially with developmental disorders, requires extra considerations related to hyperactivity, anxiety and the ability to perform and maintain attention to the fMRI paradigm. We discuss a number of methods that can be employed to minimize noise, in particular movement-related noise. To this end we focus on strategies prior to, during and following the data acquisition phase employed primarily within our own laboratory. We discuss the impact of factors such as experimental design, screening of potential participants and pre-scan training on head motion in our adolescents with developmental disorders and typical development. We make some suggestions that may minimize noise during data acquisition itself and finally we briefly discuss some current processing techniques that may help to identify and remove noise in the data. Many advances have been made in the field of pediatric imaging, particularly with regard to research involving children with developmental disorders. Mindfulness of issues such as those discussed here will ensure continued progress and greater consistency across studies.

Noise contributions to the fMRI signal: An overview

The ability to discriminate signal from noise plays a key role in the analysis and interpretation of functional magnetic resonance imaging (fMRI) measures of brain activity. Over the past two decades, a number of major sources of noise have been identified, including system-related instabilities, subject motion, and physiological fluctuations. This article reviews the characteristics of the various noise sources as well as the mechanisms through which they affect the fMRI signal. Approaches for distinguishing signal from noise and the associated challenges are also reviewed. These challenges reflect the fact that some noise sources, such as respiratory activity, are generated by the same underlying brain networks that give rise to functional signals that are of interest.

Evaluation of multi-echo ICA denoising for task based fMRI studies: Block designs, rapid event-related designs, and cardiac-gated fMRI

Multi-echo fMRI, particularly the multi-echo independent component analysis (ME-ICA) algorithm, has previously proven useful for increasing the sensitivity and reducing false positives for functional MRI (fMRI) based resting state connectivity studies. Less is known about its efficacy for task-based fMRI, especially at the single subject level. This work, which focuses exclusively on individual subject results, compares ME-ICA to single-echo fMRI and a voxel-wise T2(⁎) weighted combination of multi-echo data for task-based fMRI under the following scenarios: cardiac-gated block designs, constant repetition time (TR) block designs, and constant TR rapid event-related designs. Performance is evaluated primarily in terms of sensitivity (i.e., activation extent, activation magnitude, percent detected trials and effect size estimates) using five different tasks expected to evoke neuronal activity in a distributed set of regions. The ME-ICA algorithm significantly outperformed all other evaluated processing alternatives in all scenarios. Largest improvements were observed for the cardiac-gated dataset, where ME-ICA was able to reliably detect and remove non-neural T1 signal fluctuations caused by non-constant repetition times. Although ME-ICA also outperformed the other options in terms of percent detection of individual trials for rapid event-related experiments, only 46% of all events were detected after ME-ICA; suggesting additional improvements in sensitivity are required to reliably detect individual short event occurrences. We conclude the manuscript with a detailed evaluation of ME-ICA outcomes and a discussion of how the ME-ICA algorithm could be further improved. Overall, our results suggest that ME-ICA constitutes a versatile, powerful approach for advanced denoising of task-based fMRI, not just resting-state data.

Multilevel Dynamic Generalized Structured Component Analysis for Brain Connectivity Analysis in Functional Neuroimaging Data

We extend dynamic generalized structured component analysis (GSCA) to enhance its data-analytic capability in structural equation modeling of multi-subject time series data. Time series data of multiple subjects are typically hierarchically structured, where time points are nested within subjects who are in turn nested within a group. The proposed approach, named multilevel dynamic GSCA, accommodates the nested structure in time series data. Explicitly taking the nested structure into account, the proposed method allows investigating subject-wise variability of the loadings and path coefficients by looking at the variance estimates of the corresponding random effects, as well as fixed loadings between observed and latent variables and fixed path coefficients between latent variables. We demonstrate the effectiveness of the proposed approach by applying the method to the multi-subject functional neuroimaging data for brain connectivity analysis, where time series data-level measurements are nested within subjects.

Building a Science of Individual Differences from fMRI

To date, fMRI research has been concerned primarily with evincing generic principles of brain function through averaging data from multiple subjects. Given rapid developments in both hardware and analysis tools, the field is now poised to study fMRI-derived measures in individual subjects, and to relate these to psychological traits or genetic variations. We discuss issues of validity, reliability and statistical assessment that arise when the focus shifts to individual subjects and that are applicable also to other imaging modalities. We emphasize that individual assessment of neural function with fMRI presents specific challenges and necessitates careful consideration of anatomical and vascular between-subject variability as well as sources of within-subject variability.

Bootstrapping fMRI Data: Dealing with Misspecification

The validity of inference based on the General Linear Model (GLM) for the analysis of functional magnetic resonance imaging (fMRI) time series has recently been questioned. Bootstrap procedures that partially avoid modeling assumptions may offer a welcome solution. We empirically compare two voxelwise GLM-based bootstrap approaches: a semi-parametric approach, relying solely on a model for the expected signal; and a fully parametric bootstrap approach, requiring an additional parameterization of the temporal structure. While the fully parametric approach assumes independent whitened residuals, the semi-parametric approach relies on independent blocks of residuals. The evaluation is based on inferential properties and the potential to reproduce important data characteristics. Different noise structures and data-generating mechanisms for the signal are simulated. When the model for the noise and expected signal is correct, we find that the fully parametric approach works well, with respect to both inference and reproduction of data characteristics. However, in the presence of misspecification, the fully parametric approach can be improved with additional blocking. The semi-parametric approach performs worse than the (fully) parametric approach with respect to inference but achieves comparable results as the parametric approach with additional blocking with respect to image reproducibility. We demonstrate that when the expected signal is incorrect GLM-based bootstrapping can overcome the poor performance of classical (non-bootstrap) parametric inference. We illustrate both approaches on a study exploring the neural representation of object representation in the visual pathway.

Evaluation of sliding window correlation performance for characterizing dynamic functional connectivity and brain states

A promising recent development in the study of brain function is the dynamic analysis of resting-state functional MRI scans, which can enhance understanding of normal cognition and alterations that result from brain disorders. One widely used method of capturing the dynamics of functional connectivity is sliding window correlation (SWC). However, in the absence of a "gold standard" for comparison, evaluating the performance of the SWC in typical resting-state data is challenging. This study uses simulated networks (SNs) with known transitions to examine the effects of parameters such as window length, window offset, window type, noise, filtering, and sampling rate on the SWC performance. The SWC time course was calculated for all node pairs of each SN and then clustered using the k-means algorithm to determine how resulting brain states match known configurations and transitions in the SNs. The outcomes show that the detection of state transitions and durations in the SWC is most strongly influenced by the window length and offset, followed by noise and filtering parameters. The effect of the image sampling rate was relatively insignificant. Tapered windows provide less sensitivity to state transitions than rectangular windows, which could be the result of the sharp transitions in the SNs. Overall, the SWC gave poor estimates of correlation for each brain state. Clustering based on the SWC time course did not reliably reflect the underlying state transitions unless the window length was comparable to the state duration, highlighting the need for new adaptive window analysis techniques.

Technical Note: Independent component analysis for quality assurance in functional MRI

PURPOSE

Independent component analysis (ICA) is an established method of analyzing human functional MRI (fMRI) data. Here, an ICA-based fMRI quality control (QC) tool was developed and used.

METHODS

ICA-based fMRI QC tool to be used with a commercial phantom was developed. In an attempt to assess the performance of the tool relative to preexisting alternative tools, it was used seven weeks before and eight weeks after repair of a faulty gradient amplifier of a non-state-of-the-art MRI unit. More specifically, its performance was compared with the AAPM 100 acceptance testing and quality assurance protocol and two fMRI QC protocols, proposed by Freidman et al. ["Report on a multicenter fMRI quality assurance protocol," J. Magn. Reson. Imaging 23, 827-839 (2006)] and Stocker et al. ["Automated quality assurance routines for fMRI data applied to a multicenter study," Hum. Brain Mapp. 25, 237-246 (2005)], respectively.

RESULTS

The easily developed and applied ICA-based QC protocol provided fMRI QC indices and maps equally sensitive to fMRI instabilities with the indices and maps of other established protocols. The ICA fMRI QC indices were highly correlated with indices of other fMRI QC protocols and in some cases theoretically related to them. Three or four independent components with slow varying time series are detected under normal conditions.

CONCLUSIONS

ICA applied on phantom measurements is an easy and efficient tool for fMRI QC. Additionally, it can protect against misinterpretations of artifact components as human brain activations. Evaluating fMRI QC indices in the central region of a phantom is not always the optimal choice.

HIV infection is associated with attenuated frontostriatal intrinsic connectivity: a preliminary study

HIV-associated cognitive impairments are prevalent, and are consistent with injury to both frontal cortical and subcortical regions of the brain. The current study aimed to assess the association of HIV infection with functional connections within the frontostriatal network, circuitry hypothesized to be highly vulnerable to HIV infection. Fifteen HIV-positive and 15 demographically matched control participants underwent 6 min of resting-state functional magnetic resonance imaging (RS-fMRI). Multivariate group comparisons of age-adjusted estimates of connectivity within the frontostriatal network were derived from BOLD data for dorsolateral prefrontal cortex (DLPFC), dorsal caudate and mediodorsal thalamic regions of interest. Whole-brain comparisons of group differences in frontostriatal connectivity were conducted, as were pairwise tests of connectivity associations with measures of global cognitive functioning and clinical and immunological characteristics (nadir and current CD4 count, duration of HIV infection, plasma HIV RNA). HIV - associated reductions in connectivity were observed between the DLPFC and the dorsal caudate, particularly in younger participants (<50 years, N=9). Seropositive participants also demonstrated reductions in dorsal caudate connectivity to frontal and parietal brain regions previously demonstrated to be functionally connected to the DLPFC. Cognitive impairment, but none of the assessed clinical/immunological variables, was also associated with reduced frontostriatal connectivity. In conclusion, our data indicate that HIV is associated with attenuated intrinsic frontostriatal connectivity. Intrinsic connectivity of this network may therefore serve as a marker of the deleterious effects of HIV infection on the brain, possibly via HIV-associated dopaminergic abnormalities. These findings warrant independent replication in larger studies.

A review of fMRI simulation studies

Simulation studies that validate statistical techniques for fMRI data are challenging due to the complexity of the data. Therefore, it is not surprising that no common data generating process is available (i.e. several models can be found to model BOLD activation and noise). Based on a literature search, a database of simulation studies was compiled. The information in this database was analysed and critically evaluated focusing on the parameters in the simulation design, the adopted model to generate fMRI data, and on how the simulation studies are reported. Our literature analysis demonstrates that many fMRI simulation studies do not report a thorough experimental design and almost consistently ignore crucial knowledge on how fMRI data are acquired. Advice is provided on how the quality of fMRI simulation studies can be improved.

Scanning fast and slow: current limitations of 3 Tesla functional MRI and future potential

Functional MRI at 3T has become a workhorse for the neurosciences, e.g., neurology, psychology, and psychiatry, enabling non-invasive investigation of brain function and connectivity. However, BOLD-based fMRI is a rather indirect measure of brain function, confounded by physiology related signals, e.g., head or brain motion, brain pulsation, blood flow, intermixed with susceptibility differences close or distant to the region of neuronal activity. Even though a plethora of preprocessing strategies have been published to address these confounds, their efficiency is still under discussion. In particular, physiological signal fluctuations closely related to brain supply may mask BOLD signal changes related to "true" neuronal activation. Here we explore recent technical and methodological advancements aimed at disentangling the various components, employing fast multiband vs. standard EPI, in combination with fast temporal ICA. Our preliminary results indicate that fast (TR <0.5 s) scanning may help to identify and eliminate physiologic components, increasing tSNR and functional contrast. In addition, biological variability can be studied and task performance better correlated to other measures. This should increase specificity and reliability in fMRI studies. Furthermore, physiological signal changes during scanning may then be recognized as a source of information rather than a nuisance. As we are currently still undersampling the complexity of the brain, even at a rather coarse macroscopic level, we should be very cautious in the interpretation of neuroscientific findings, in particular when comparing different groups (e.g., age, sex, medication, pathology, etc.). From a technical point of view our goal should be to sample brain activity at layer specific resolution with low TR, covering as much of the brain as possible without violating SAR limits. We hope to stimulate discussion toward a better understanding and a more quantitative use of fMRI.