fMRI Noise Research Articles

Improving laminar fMRI specificity by reducing macrovascular bias revealed by respiration effects

Abstract Functional MRI (fMRI) time series are inherently susceptible to the influence of respiratory variations. While many studies treat respiration as a source of noise in fMRI, this study employs natural respiratory variations during high resolution (0.8mm) fMRI at 7T to formulate a respiration effect related map and then use this map to reduce macrovascular bias for a more laminar-specific fMRI measurement. Our results indicate that respiratory related signal changes are modulated by breath phase (breathing in/out or in the transition between breath in and out) during fMRI acquisition, with distinct patterns across various brain regions. We demonstrate that respiration maps generated from normal fMRI runs, such as task-oriented sessions, closely resemble those from deep-breath and breath-hold experiments. These maps show a significant correlation with the macro-vasculature automatically segmented based on SWI and QSM images. Most crucially, by removing voxels most responsive to respiratory variations, we can refine high-resolution fMRI measurements to be more layer-specific, improving the accuracy of laminar fMRI analysis.

Web-based processing of physiological noise in fMRI: addition of the PhysIO toolbox to CBRAIN.

Neuroimaging research requires sophisticated tools for analyzing complex data, but efficiently leveraging these tools can be a major challenge, especially on large datasets. CBRAIN is a web-based platform designed to simplify the use and accessibility of neuroimaging research tools for large-scale, collaborative studies. In this paper, we describe how CBRAIN's unique features and infrastructure were leveraged to integrate TAPAS PhysIO, an open-source MATLAB toolbox for physiological noise modeling in fMRI data. This case study highlights three key elements of CBRAIN's infrastructure that enable streamlined, multimodal tool integration: a user-friendly GUI, a Brain Imaging Data Structure (BIDS) data-entry schema, and convenient in-browser visualization of results. By incorporating PhysIO into CBRAIN, we achieved significant improvements in the speed, ease of use, and scalability of physiological preprocessing. Researchers now have access to a uniform and intuitive interface for analyzing data, which facilitates remote and collaborative evaluation of results. With these improvements, CBRAIN aims to become an essential open-science tool for integrative neuroimaging research, supporting FAIR principles and enabling efficient workflows for complex analysis pipelines.

Resting state network connectivity is attenuated by fMRI acoustic noise

IntroductionDuring the past decades there has been an increasing interest in tracking brain network fluctuations in health and disease by means of resting state functional magnetic resonance imaging (rs-fMRI). Rs-fMRI however does not provide the ideal environmental setting, as participants are continuously exposed to noise generated by MRI coils during acquisition of Echo Planar Imaging (EPI). We investigated the effect of EPI noise on resting state activity and connectivity using magnetoencephalography (MEG), by reproducing the acoustic characteristics of rs-fMRI environment during the recordings. As compared to fMRI, MEG has little sensitivity to brain activity generated in deep brain structures, but has the advantage to capture both the dynamic of cortical magnetic oscillations with high temporal resolution and the slow magnetic fluctuations highly correlated with BOLD signal. MethodsThirty healthy subjects were enrolled in a counterbalanced design study including three conditions: a) silent resting state (Silence), b) resting state upon EPI noise (fMRI), and c) resting state upon white noise (White). White noise was employed to test the specificity of fMRI noise effect. The amplitude envelope correlation (AEC) in alpha band measured the connectivity of seven Resting State Networks (RSN) of interest (default mode network, dorsal attention network, language, left and right auditory and left and right sensory-motor). Vigilance dynamic was estimated from power spectral activity. ResultsfMRI and White acoustic noise consistently reduced connectivity of cortical networks. The effects were widespread, but noise and network specificities were also present. For fMRI noise, decreased connectivity was found in the right auditory and sensory-motor networks. Progressive increase of slow theta-delta activity related to drowsiness was found in all conditions, but was significantly higher for fMRI . Theta-delta significantly and positively correlated with variations of cortical connectivity. Discussionrs-fMRI connectivity is biased by unavoidable environmental factors during scanning, which warrant more careful control and improved experimental designs. MEG is free from acoustic noise and allows a sensitive estimation of resting state connectivity in cortical areas. Although underutilized, MEG could overcome issues related to noise during fMRI, in particular when investigation of motor and auditory networks is needed.

Effects of acoustic fMRI-noise on taste identification, liking, and intensity

• First study to compare gustatory perception in a silent and fMRI-noise setting. • Participants underwent gustatory testing in a silent and fMRI-noise setting in randomised groups. • No significant effect of loud acoustic fMRI noise on taste perception (liking, intensity, or identification) was found. • fMRI noise does not seem to constitute a systematic confounder in gustatory research. • Further research should be conducted with regards to other potential confounders. In this paper, we investigated the possible effect of loud, acoustic noise produced by an fMRI-scanner on gustatory perception compared to a silent setting. The purpose was to assess whether the loud acoustic noise produced during fMRI acquisition could constitute a systematic confounder in most human gustatory research. To date, this issue has not yet been addressed. To evaluate this, fifty participants were tested in a randomised order in an fMRI-noise setting and a quiet setting, respectively. We found no significant effect of loud, acoustic fMRI noise on taste perception, including identification, certainty of identification, perceived intensity and hedonic rating. Thus, the profound acoustic commotion experienced by subjects undergoing an fMRI does not seem to constitute a systematic confounder in gustatory research.

FMRI Acoustic Noise Enhances Parasympathetic Activity in Humans.

Background: Functional magnetic resonance imaging (fMRI) is one of the most important neuroimaging techniques; nevertheless, the acoustic noise of the MR scanner is unavoidably linked to the process of data acquisition. We hypothesized that the auditory noise of the scanner has an effect on autonomic activity. Methods: We measured heart rate variability (HRV) while exposing 30 healthy subjects to fMRI noise. In doing so, we demonstrated an increase in parasympathetic nervous system (PNS) activity compared to silence and white noise and a decrease in sympathetic nervous system (SNS) activity compared to white noise. Conclusions: The influence of MR scanner noise on the autonomic nervous system should be taken into account when performing fMRI experiments.

Reconstruction of natural images from evoked brain activity with a dictionary-based invertible encoding procedure

Studies on visual encoding and reconstruction based on functional magnetic resonance imaging (fMRI) have inspired each other in recent years. However, as far as we know, there has not been any study that has achieved the reconstruction of natural stimuli by directly reversing an encoding model with strong interpretability; in other words, the interpretability of current decoding methods is weak. To solve this problem, we first design a reversible feature extraction method using Gabor wavelets and build a nonnegative sparse mapping between the features and brain activity, thus achieving visual encoding. Then, based on the mapping, we estimate the features from measured brain activity and reverse the feature extraction method step-by-step. In this process, we use dictionary-learning technology to explore the natural statistical structure of the features from the image database, thereby greatly reducing the negative impact of information loss and fMRI noise. Finally, the stimuli can be reconstructed from the estimated features by back-propagation. Because the encoding procedure is highly transparent, the reconstruction procedure obtained by reversing the encoding model is also highly interpretable. The experiments show that our encoding method can build effective voxel-wise models for early visual areas, and also show that the proposed method is capable of reconstructing the basic outline of the stimuli with low structural complexity.

Resting-state fMRI noise regression in multi-site aging studies

Resting-state fMRI noise regression in multi-site aging studies

Attenuation of motion artifacts in fMRI using discrete reconstruction of irregular fMRI trajectories (DRIFT).

Numerous studies report motion as the most detrimental source of noise and artifacts in fMRI. Current motion correction methods fail to completely address the motion problem. Retrospective techniques such as spatial realignment can correct for between-volume misalignment but fail to address within volume contamination and spin-history artifacts. Prospective motion correction can prevent spin-history artifacts but currently cannot update the gradients fast enough to remove k-space filling artifacts, calling for a hybrid approach to fully address these problems. Motion can be mathematically formulated into the MR signal equation to describe the motion artifacts at their origin in k-space. From these equations, it is demonstrated that different motions have different effects on the signal. A novel motion correction algorithm is designed from these equations to remove motion-induced artifacts directly in k-space, discrete reconstruction of irregular fMRI trajectory (DRIFT). This method is evaluated rigorously using fMRI simulations and data from a rotating phantom inside the scanner. The results indicate that although some motion types have negligible effects on the MR signal, others produce catastrophic and lasting artifacts even after motion cessation. In simulation, DRIFT is able to remove motion artifacts in the absence of spin history. In a phantom scan, DRIFT significantly attenuates the motion artifacts in the fMRI data. Neither prospective nor retrospective motion correction methods could completely remove the motion artifacts from the fMRI data. However, DRIFT, as a retrospective technique, when combined with prospective motion correction, can eliminate a significant portion of motion artifacts.

Sailing in rough waters: Examining volatility of fMRI noise

BackgroundThe assumption that functional magnetic resonance imaging (fMRI) noise has constant volatility has recently been challenged by studies examining heteroscedasticity arising from head motion and physiological noise. The present study builds on this work using latest methods from the field of financial mathematics to model fMRI noise volatility. MethodsMulti-echo phantom and human fMRI scans were used and realised volatility was estimated. The Hurst parameter H ∈ (0,1), which governs the roughness/irregularity of realised volatility time series, was estimated. Calibration of H was performed pathwise, using well-established neural network calibration tools. ResultsIn all experiments the volatility calibrated to values within the rough case, H < 0.5, and on average fMRI noise was very rough with 0.03 < H < 0.05. Some edge effects were also observed, whereby H was larger near the edges of the phantoms. DiscussionThe findings suggest that fMRI volatility is not only non-constant, but also substantially more irregular than a standard Brownian motion. Thus, further research is needed to examine the impact such pronounced oscillations in the volatility of fMRI noise have on data analyses.

A novel spatiotemporal tool for the automatic classification of fMRI noise based on Independent Component Analysis.

In this study, a semi-automatic, easy-to-use classification method for the identification and removal of fMRI noise is proposed and tested. The method relies on subject-level spatial independent component analysis (ICA) of fMRI data. Starting from a reference set of labeled independent components (ICs), novel ICs are classified as physiological/artefactual by combining a spatial correlation (SC) analysis with the reference ICs and relative power spectral (PS) analysis. Here, ICs from a task-based fMRI dataset were used as reference. SC and SP thresholds were set using a test dataset (5 subjects, same fMRI protocol) based on Receiving Operating Characteristic curves. The tool performance and versatility were measured on a resting-state fMRI dataset (5 subjects). Our results show that the method can automatically identify noise-related ICs with accuracy, specificity and sensitivity higher than 80% across different fMRI protocols. These findings also suggest that the reference set provided in the present study might be used to mark ICs coming from independent taskrelated or resting-state fMRI datasets.Clinical relevance- The new method will be included in a userfriendly, open-source tool for removal of noisy contributions from fMRI datasets to be used in clinical and research practices.

Statistical analysis of fMRI using wavelets: Big Data, denoising, large‐p‐small‐n matrices

AbstractOver the past decade functional Magnetic Resonance Imaging (fMRI) has been intensively used to study the complex functional network organization of the human brain and how it changes in time. An fMRI machine produces 3D time‐course cerebral images that contain hundreds of thousands of voxels and each voxel is scanned for hundreds of times. This potentially allows the researchers to explore functional connectivity on a voxel‐to‐voxel level, and also yields a number of serious statistical complications. First of all, the high‐dimension property of fMRI data turns it into Big Data. Furthermore, the study of functional brain network for so many voxels involves the problem of estimation of and simultaneous inference for large‐p‐small‐n cross‐covariance matrices. Furthermore, all problems should be solved in the presence of notoriously large fMRI noise which often forces statisticians to average signals over large areas instead of considering a network between individual voxels. An attractive alternative to the averaging, discussed in the paper, is a multiresolution wavelet analysis complemented by special procedures of estimating noise and estimation and simultaneous inference for cross‐covariance and cross‐correlation matrices for hundreds of thousands pairs of voxels, and it is fair to say that if wavelets have not been already known, fMRI applications would necessitates their creation. Both task and resting‐state fMRI are considered, and lessons from the wavelet analysis of ultra‐fast and conventional neuroplasticity fMRI experiments are presented. The article is self‐contained and does not require familiarity with wavelets or fMRI.This article is categorized under: Algorithms and Computational Methods &gt; Numerical Methods Applications of Computational Statistics &gt; Signal and Image Processing and Coding

Dynamic-flip-angle ECG-gating with nuisance signal regression improves resting-state BOLD functional connectivity mapping by reducing cardiogenic noise.

To investigate an ECG-gated dynamic-flip-angle BOLD sequence with improved robustness against cardiogenic noise in resting-state fMRI. ECG-gating minimizes the cardiogenic noise but introduces T1 -dependent signal variation, which is minimized by combination of a dynamic-flip-angle technique and retrospective nuisance signal regression (NSR) using signals of white matter, CSF, and global average. The technique was studied with simulations in a wide range of T1 and B1 fields and phantom imaging with pre-programmed TR variations. Resting-state fMRI of 20 healthy subjects was acquired with non-gated BOLD (NG), ECG-gated constant-flip-angle BOLD (GCFA), ECG-gated BOLD with retrospective T1 -correction (GRC), and ECG-gated dynamic-flip-angle BOLD (GDFA), all processed by the same NSR method. GDFA was compared to alternative methods over temporal SNR (tSNR), seed-based connectivity, and whole-brain voxelwise connectivity based on intrinsic connectivity distribution (ICD). A previous large-cohort data set (N = 100) was used as a connectivity gold standard. Simulations and phantom imaging show substantial reduction of the T1 -dependent signal variation with GDFA alone, and further reduction with NSR. The resting-state study shows improved tSNR in the basal brain, comparing GDFA to NG, after both processed with NSR. Furthermore, GDFA significantly improved subcortical-subcortical and cortical-subcortical connectivity for several representative seeds and significantly improved ICD in the brainstem, thalamus, striatum, and prefrontal cortex, compared to the other 3 approaches. GDFA with NSR improves mapping of the resting-state functional connectivity of the basal-brain regions by reducing cardiogenic noise.

Wavelet Analysis of Big Data Contaminated by Large Noise in an fMRI Study of Neuroplasticity

Functional magnetic resonance imaging (fMRI) allows researchers to analyze brain activity on a voxel level, but using this ability is complicated by dealing with Big Data and large noise. A traditional remedy is averaging over large parts of brain in combination with more advanced technical innovations in reducing fMRI noise. In this paper a novel statistical approach, based on a wavelet analysis of standard fMRI data, is proposed and its application to an fMRI study of neuron plasticity of 24 healthy adults is presented. The aim of that study was to recognize changes in connectivity between left and right motor cortices (the neuroplasticity) after button clicking training sessions. A conventional method of the data analysis, based on averaging images, has implied that for the group of 24 participants the connectivity increased after the training. The proposed wavelet analysis suggests to analyze pathways between left and right hemispheres on a voxel-to-voxel level and for each participant via estimation of corresponding cross-correlations. This immediately necessitates statistical analysis of large-p-small-n correlation matrices contaminated by large noise. Furthermore, distributions that we are dealing in the analysis are neither Gaussian nor sub-Gaussian but sub-exponential. The paper explains how the problem may be solved and presents results of a dynamic analysis of the ability of a human brain to reorganize itself for 24 healthy adults. Results show that the ability of a brain to reorganize itself varies widely even among healthy individuals, and this observation is important for our understanding of a human brain and treatment of brain diseases.

Impacting the effect of fMRI noise through hardware and acquisition choices – Implications for controlling false positive rates

We review the components of time-series noise in fMRI experiments and the effect of image acquisition parameters on the noise. In addition to helping determine the total amount of signal and noise (and thus temporal SNR), the acquisition parameters have been shown to be critical in determining the ratio of thermal to physiological induced noise components in the time series. Although limited attention has been given to this latter metric, we show that it determines the degree of spatial correlations seen in the time-series noise. The spatially correlations of the physiological noise component are well known, but recent studies have shown that they can lead to a higher than expected false-positive rate in cluster-wise inference based on parametric statistical methods used by many researchers. Based on understanding the effect of acquisition parameters on the noise mixture, we propose several acquisition strategies that might be helpful reducing this elevated false-positive rate, such as moving to high spatial resolution or using highly-accelerated acquisitions where thermal sources dominate. We suggest that the spatial noise correlations at the root of the inflated false-positive rate problem can be limited with these strategies, and the well-behaved spatial auto-correlation functions (ACFs) assumed by the conventional statistical methods are retained if the high resolution data is smoothed to conventional resolutions.

Effect of temporal resolution and serial autocorrelations in event‐related functional MRI

To assess the impact of colored noise on statistics in event-related functional MRI (fMRI) (visual stimulation using checkerboards) acquired by simultaneous multislice imaging enabling repetition times (TRs) between 2.64 to 0.26 s. T-values within the visual cortex obtained with analysis tools that assume a first-order autoregressive plus white noise process (AR(1)+w) with a fixed AR coefficient versus higher-order AR models with spatially varying AR coefficients were compared. In addition, dependency of T-values on correction of physiological noise (respiration, heart rate) was evaluated. Optimal statistical power was obtained for a TR of 0.33 s, but T-values as obtained by AR(1)+w models were strongly dependent on the predefined AR coefficients in fMRI with short TRs which required higher-order AR models to achieve stable statistics. Direct estimation of AR coefficients revealed the highest values within the default mode network while physiological noise had little influence on statistics in cortical structures. Colored noise in event-related fMRI obtained at short TRs originates mainly from neural sources and calls for more sophisticated correction of serial autocorrelations which cannot be achieved with standard methods relying on AR(1)+w models with globally fixed AR coefficients. Magn Reson Med 76:1805-1813, 2016. © 2016 International Society for Magnetic Resonance in Medicine.

Physiological noise in human cerebellar fMRI

To compare physiological noise contributions in cerebellar and cerebral regions of interest in high-resolution functional magnetic resonance imaging (fMRI) data acquired at 7T, to estimate the need for physiological noise removal in cerebellar fMRI. Signal fluctuations in high resolution (1mm isotropic) 7T fMRI data were attributed to one of the following categories: task-induced BOLD changes, slow drift, signal changes correlated with the cardiac and respiratory cycles, signal changes related to the cardiac rate and respiratory volume per unit of time or other. [Formula: see text] values for all categories were compared across regions of interest. In this high-resolution data, signal fluctuations related to the phase of the cardiac cycle and cardiac rate were shown to be significant, but comparable between cerebellar and cerebral regions of interest. However, respiratory related signal fluctuations were increased in the cerebellar regions, with explained variances that were up to 80% higher than for the primary motor cortex region. Even at a millimetre spatial resolution, significant correlations with both cardiac and respiratory RETROICOR components were found in all healthy volunteer data. Therefore, physiological noise correction is highly likely to improve the temporal signal-to-noise ratio (SNR) for cerebellar fMRI at 7T, even at high spatial resolution.

Brain activation inhomogeneity highlighted by the Isotropic Anomalous Diffusion filter.

The visual appealing nature of the now popular BOLD fMRI may give the false impression of extreme simplicity, as if the the functional maps could be generated with the press of a single button. However, one can only get plausible maps after long and cautious processing, considering that time and noise come into play during acquisition. One of the most popular ways to account for noise and individual variability in fMRI is the use of a Gaussian spatial filter. Although very robust, this filter may introduce excessive blurring, given the strong dependence of results on the central voxel value. Here, we propose the use of the Isotropic Anomalous Diffusion (IAD) approach, aiming to reduce excessive homogeneity while retaining the natural variability of signal across brain space. We found differences between Gaussian and IAD filters in two parameters gathered from Independent Component maps (ICA), identified on brain areas responsible for auditory processing during rest. Analysis of data gathered from 7 control subjects shows that the IAD filter rendered more localized active areas and higher contrast-to-noise ratios, when compared to equivalent Gaussian filtered data (Student t-test, p<0.05). The results seem promising, since the anomalous filter performs satisfactorily in filtering noise with less distortion of individual localized brain responses.

Highcor: A novel data-driven regressor identification method for BOLD fMRI

A data-driven regressor selection technique for reducing physiological noise in BOLD fMRI is presented that capitalizes on additional information contained in the phase of the signal time-course. This method, termed highcor, identifies a set of suspect voxels by selecting based on high temporal correlation between the magnitude and phase components of the time-course. Temporal regressors are generated from principal component analysis of this voxel set. Regressor spectral content is investigated with high temporal resolution datasets, and filtering performance is demonstrated. The technique is benchmarked against compcor, an increasingly popular data driven technique. Highcor was found to select a unique set of physiological noise source voxels, and identification of confound and physiologically related signals was robust even at slow temporal sampling rates. Filtering using regressors derived from compcor and highcor voxels resulted in reductions in overall temporal standard deviation in cortical areas of 16.1%±3.1%, and 18.1%±3.8%, (mean±sd.) as measured over 36 BOLD fMRI datasets that featured an anti-saccade task. An approach combining both methods resulted in further reductions of temporal standard deviations by 31.4%±3.8%. In these regions, mean temporal SNR values were improved from 38.7±3.4 to 47.7±3.7 (cc), 49.2±2.2 (hc), and 57.8±2.3 (hc+cc). tSNR increases from data-driven filtering translated into some associated improvements in overall detection of task in the sample datasets.

On the impact of physiological noise in spinal cord functional MRI.

Functional magnetic resonance imaging (fMRI) techniques are widely exploited for the study of brain activation. In recent years, similar approaches have been attempted for the study of spinal cord function; however, obtaining good functional images of spinal cord still represents a technical and scientific challenge. Some of the main limiting factors can be classified under the broad category of "physiological noise," and are related to 1) the cerebrospinal fluid (CSF) flux in the subarachnoid space surrounding the spinal cord; 2) the cord motion itself; and 3) the small area of the cord, which makes it critical to have a high image resolution. In addition, the different magnetic susceptibility properties of tissues surrounding the spinal cord reduce the local homogeneity of the static magnetic field, causing image distortion, reduction of the effective resolution, and signal loss, all effects that are modulated by motion. For these reasons, a number of methods have been developed for the purpose of denoising spinal cord fMRI time series. In this work, after a short introduction on the relevant features of the spinal cord anatomy, we review the main sources of physiological noise in spinal cord fMRI and discuss the main approaches useful for its mitigation.

Physiological noise in cerebellar fMRI

Physiological noise in cerebellar fMRI