Physiological Noise Research Articles

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.

Selective attention reduces physiological noise in the external ear canals of humans. II: Visual attention

Human subjects performed in several behavioral conditions requiring, or not requiring, selective attention to visual stimuli. Specifically, the attentional task was to recognize strings of digits that had been presented visually. A nonlinear version of the stimulus-frequency otoacoustic emission (SFOAE), called the nSFOAE, was collected during the visual presentation of the digits. The segment of the physiological response discussed here occurred during brief silent periods immediately following the SFOAE-evoking stimuli. For all subjects tested, the physiological-noise magnitudes were substantially weaker (less noisy) during the tasks requiring the most visual attention. Effect sizes for the differences were >2.0. Our interpretation is that cortico-olivo influences adjusted the magnitude of efferent activation during the SFOAE-evoking stimulation depending upon the attention task in effect, and then that magnitude of efferent activation persisted throughout the silent period where it also modulated the physiological noise present. Because the results were highly similar to those obtained when the behavioral conditions involved auditory attention, similar mechanisms appear to operate both across modalities and within modalities. Supplementary measurements revealed that the efferent activation was spectrally global, as it was for auditory attention.

Selective attention reduces physiological noise in the external ear canals of humans. I: Auditory attention

In this study, a nonlinear version of the stimulus-frequency OAE (SFOAE), called the nSFOAE, was used to measure cochlear responses from human subjects while they simultaneously performed behavioral tasks requiring, or not requiring, selective auditory attention. Appended to each stimulus presentation, and included in the calculation of each nSFOAE response, was a 30-ms silent period that was used to estimate the level of the inherent physiological noise in the ear canals of our subjects during each behavioral condition. Physiological-noise magnitudes were higher (noisier) for all subjects in the inattention task, and lower (quieter) in the selective auditory-attention tasks. These noise measures initially were made at the frequency of our nSFOAE probe tone (4.0 kHz), but the same attention effects also were observed across a wide range of frequencies. We attribute the observed differences in physiological-noise magnitudes between the inattention and attention conditions to different levels of efferent activation associated with the differing attentional demands of the behavioral tasks. One hypothesis is that when the attentional demand is relatively great, efferent activation is relatively high, and a decrease in the gain of the cochlear amplifier leads to lower-amplitude cochlear activity, and thus a smaller measure of noise from the ear.

31. Simulations of fMRI data with usage of estimated physiological artifacts

Introduction Simulation of fMRI data is a useful tool to verify or to compare different processing methods. Typically simulated data are created using combination of modeled signals and regions of interest with addition of some noise (Gaussian, sometimes low-frequency noise, autocorrelation or polynomial drift). Our previous experiences indicated that such simplification is not enough to simulate 4D data suitable for independent component analysis (ICA). In this work, we introduced new approach which uses estimates of physiological artifacts from real data to create more realistic structure of simulated fMRI data. Methods Synthetic fMRI datasets were created using in-house fMRI simulator. We combined three regions of interest sharing the same signal fluctuation according to block stimulation, three other regions with low-frequency fluctuations and random events uncorrelated to stimulation, EPI template, Gaussian and low-frequency noise, and polynomial drift. RETROICOR algorithm was used to estimate spatial maps and temporal characteristics of cardiac and breathing processes from resting-state fMRI data. These estimates of physiological artifacts were combined with synthetic data in multiplicative manner to form the final 4D datasets. ICA was used to process both (1) pure synthetic datasets, and (2) datasets with physiological noise to compare the suitability of our process. Results Typically 4 components were estimated using ICA on pure synthetic data showing all modeled regions. Using more realistic datasets, typically 7 additional components related to physiological noise were estimated. Conclusion Using more realistic simulated data will enable us, for instance, to test classification algorithm and to make ICA calculation more realistic because using ICA with resting-state data results typically in 20 components (and more than half are probably related to artifacts). We proved that using spatiotemporal characteristics of physiological processes estimated from real data, it is possible to create more realistic simulations of fMRI data.

23. Statistical characteristics of event related and block design datasets

In this work we characterize functional magnetic resonance imaging (fMRI) data from event related and block design studies with typical values of several parameters. We use signal to noise ratio (SNR), percent signal change (PSC) and statistical parametric maps, estimated by SPM8 toolbox (beta, con and t-statistic maps). The aim of this work is to describe typical values appearing in datasets and to enhance knowledge about typical distributions of dataset’s parameters which could be used for estimation of data quality. We used data from four studies. Datasets were acquired using 1.5T MR scanner Siemens Symphony. First dataset is event-related visual oddball experiment (VOB) with 4 sessions, other datasets are block design studies: visual memory task (VMT), verbal fluency task (VFT) and semantic decision task (SDT). Parameters were calculated for three types of previously selected brain regions – regions related to the task (activations and deactivations), regions related to physiological artifacts and one cortical area not related to the task. In the event related study, typical PSC in task activated voxels was about 0.1%, in the block design studies was PSC higher, with mean values between 0.4% and 1%. Greater variance of SNR was observed in regions designated as sources of physiological noise in event related than in block design studies. We also observed significant correlation between t -values and PSC in VOB study (Pearson p = 0.8517) and stronger correlations in block design studies (Pearson p = {0.9902; 0.9299; 0.9697}). Lower correlation of PSC and t -values in VOB study could be caused due to separation of experiment in four sessions. More robust explanation could be made after including more event related studies into this comparison. These dataset characterizations will provide useful basis for fMRI data simulations, for calculation of statistical power, and possibly for evaluation of fMRI data quality. Thanks to Grant P103/12/0552 from GACR for funding.

Measurement of hearing-protector attenuation using auditory steady state responses

There is a need to assess hearing protection device (HPD) attenuation to ensure that individuals receive adequate protection from noise. Present methods of attenuation measurement have limitations. Objective measurements such as field microphone in real ear (F-MIRE) do not assess bone conducted sound. Psychophysical measurements such as real ear attenuation at threshold (REAT) are biased due to the low frequency masking effects from test subjects' physiological noise, and the variability of measurements based on subjective response. We explored using auditory steady state responses (ASSR) as a technique that may overcome these limitations. ASSRs were recorded in ten normal hearing adults, using both “normal” and “occluded” conditions. Stimuli included both narrow band noises and pure tones (500 and 1000 Hz), amplitude modulated at 40 Hz. Stimuli were presented through either loudspeakers or headphones, at 45, 55, and 65 dB SPL. “Physiological attenuation” was calculated as the difference between ASSR values (both amplitude and phase) for normal and occluded conditions. Physiological attenuation estimates were compared to in-ear (objective) and psychophysical (subjective) measurements. Grand mean ASSR data complied well with in-ear and subjective measurements. Further work is needed to provide accurate assessment of ASSR-based physiological attenuation for individual subjects.

The effects of age on multiscale entropy measurement of neurophysiological signal complexity (1125.6)

Complex physiological signals are neither deterministic nor random, showing a mixture of long‐range dependencies and irregularities (i.e., information) over multiple time scales. In humans, aging and pathology are associated with diminished CNS neurophysiological complexity and impaired cognition. In this study, we computed multiscale entropy (MSE) to measure the time scale dependent complexity (i.e., information) of the resting state BOLD fMRI signal in humans over a large age span (N=138; ages 8‐83; 58 male). Physiological noise (heart rate, respiration, head motion and local white matter [WM] signals) was removed from fMRI time courses using linear regression. Data were then bandpass filtered (.008‐.1 Hz) to isolate low frequency fluctuations. MSE was computed with parameters m=2 and r=.2 for temporal scales 1‐18. BOLD signal in gray matter (GM) and WM gained information with increasing temporal scale and exceeded the information gain from bandpass filtered white and 1/f noise. GM sustained the information over coarse scales better than WM. Age groups stratified MSE profiles showing an age‐related loss of signal complexity in GM that was maximal at scale 18. In contrast, MSE from WM did not dissociate age groups at any scale. These results show that BOLD fMRI is a complex physiological signal characterized by a tissue‐specific age‐related reduction in complexity.

RETROSPECTIVE DETECTION OF INTERLEAVED SLICE ACQUISITION PARAMETERS FROM FMRI DATA.

To minimize slice excitation leakage to adjacent slices, interleaved slice acquisition is nowadays performed regularly in fMRI scanners. In interleaved slice acquisition, the number of slices skipped between two consecutive slice acquisitions is often referred to as the 'interleave parameter'; the loss of this parameter can be catastrophic for the analysis of fMRI data. In this article we present a method to retrospectively detect the interleave parameter and the axis in which it is applied. Our method relies on the smoothness of the temporal-distance correlation function, which becomes disrupted along the axis on which interleaved slice acquisition is applied. We examined this method on simulated and real data in the presence of fMRI artifacts such as physiological noise, motion, etc. We also examined the reliability of this method in detecting different types of interleave parameters and demonstrated an accuracy of about 94% in more than 1000 real fMRI scans.

Effects of self-generated noise on estimates of detection threshold in quiet in school-age children and adults

Detection in quiet develops earlier in childhood for high than low frequencies. The present study tested the hypothesis that self-generated noise could play a role in this finding. When adults listen for sounds near threshold, they tend to engage in behaviors that reduce physiologic noise (e.g., quiet breathing), which is predominantly low frequency. Children may not suppress self-generated noise to the same extent as adults. This possibility was evaluated by measuring sound levels in the ear canal simultaneous with adaptive threshold estimation for 250-, 1000-, and 4000-Hz pure tones. Stimuli were delivered and recordings were made using a single foam insert. Listeners were children (4.3–16.0 yr) or adults. Consistent with previous data, the effect of child age was robust at 250 Hz, whereas thresholds of even the youngest listeners were nearly adult-like at 4000 Hz. The spectral shape of self-generated noise was generally similar across listener age groups, although the magnitude was higher in younger listeners. Trial-by-trial data were evaluated to assess the relationship between noise and the accuracy of listener responses: there was an association for younger listeners. These results provide preliminary evidence that self-generated noise may play a role in the prolonged development of low-frequency detection in quiet.

Studying the Spatial Distribution of Physiological Effects on BOLD Signals Using Ultrafast fMRI.

The blood-oxygen-level dependent (BOLD) signal in functional MRI (fMRI) reflects both neuronal activations and global physiological fluctuations. These physiological fluctuations can be attributed to physiological low frequency oscillations (pLFOs), respiration, and cardiac pulsation. With typical TR values, i.e., 2 s or longer, the high frequency physiological signals (i.e., from respiration and cardiac pulsation) are aliased into the low frequency band, making it hard to study the individual effect of these physiological processes on BOLD. Recently developed multiband EPI sequences, which offer full brain coverage with extremely short TR values (400 ms or less) allow these physiological signals to be spectrally separated. In this study, we applied multiband resting state scans on nine healthy participants with TR = 0.4 s. The spatial distribution of each physiological process on BOLD fMRI was explored using their spectral features and independent component analysis (ICA). We found that the spatial distributions of different physiological processes are distinct. First, cardiac pulsation affects mostly the base of the brain, where high density of arteries exists. Second, respiration affects prefrontal and occipital areas, suggesting the motion associated with breathing might contribute to the noise. Finally, and most importantly, we found that the effects of pLFOs dominated many prominent ICA components, which suggests that, contrary to the popular belief that aliased cardiac and respiration signals are the main physiological noise source in BOLD fMRI, pLFOs may be the most influential physiological signals. Understanding and measuring these pLFOs are important for denoising and accurately modeling BOLD signals.

0411: Improved spatial resolution and post-processing for myocardial blood flow quantification in humans using steady-pulsed arterial spin labeling

Arterial spin labeling (ASL) is a powerful tool for the noninvasive assessment of tissue perfusion. In the human heart, however, measuring myocardial perfusion (MBF) is challenging due to strong physiological noise. Steady-pulsed ASL (spASL) under free-breathing had been proposed to improve sensitivity. To improve robustness against respiratory motion, efficiency and spatial resolution, we present an optimization of the postprocessing algorithm by way of a dedicated motion correction (Moco). The spASL sequence was implemented on a Siemens 3T Verio system, based on an ECG-triggered bSSFP acquisition combined with a sliceselective labeling module. ASL was performed at each cardiac cycle to drive tissue magnetization into a perfusion-dependent steady-state. The spASL acquisition was repeated 128 times to accumulate data and a customized 2-step Moco was carried out. First, global correction in a large ROI was performed by spatially shifting every image and minimizing signal difference with a reference. Then, a more precise regional correction based on contour correlation was used. When performing global correction, MBF mapping was possible. Fig.1 shows a comparison between two maps in a volunteer without and with Moco. Better homogeneity and delineation of the myocardium can be seen. Using the 2nd described step before carrying out the quantification in myocardial regions led to improved signal stability. Without Moco, signal was found stable when including up to 30% of all images. With Moco, the signal reached a plateau around 50% and remained stable until 80% of included images. With this new approach, a greater portion (80% versus 30% formerly) of the acquired data can be used for regional perfusion assessment. This work is also a step towards calculation of myocardial perfusion maps with high spatial resolution using ASL. Abstract 0411 – Figure: MBF maps obtained with spASL MRI.

The effect of echo time and post-processing procedure on blood oxygenation level-dependent (BOLD) functional connectivity analysis

While spontaneous BOLD fMRI signal is a common tool to map functional connectivity, unexplained inter- and intra-subject variability frequently complicates interpretation. Similar to evoked BOLD fMRI responses, spontaneous BOLD signal is expected to vary with echo time (TE) and corresponding intra/extravascular sensitivity. This may contribute to discrepant conclusions even following identical post-processing pipelines. Here we applied commonly-utilized independent component analysis (ICA) as well as seed-based correlation analysis and investigated default mode network (DMN) and visual network (VN) detection from BOLD data acquired at three TEs (3T; TR=2500ms; TE=15ms, 35ms, and 55ms) and from quantitative R2* maps. Explained variance in ICA analysis was significantly higher (P<0.05) when R2*-derived maps were considered relative to single-TE data with no post-processing. While explained variance in the BOLD data increased with motion correction, R2* derived DMN and VN were minimally affected by motion correction. Explained variance increased in all data when physiological noise confounds were removed using CompCor. Notably, the R2*-derived connectivity patterns were least affected by motion and physiological noise confounds in a seed-based correlation analysis. Intermediate (35ms) and long (55ms) TE data provided similar spatial and temporal characteristics only after reducing motion and physiological noise contamination. Results provide an exemplar for how 3T spontaneous BOLD network detection varies with TE and post-processing procedure over the range of commonly acquired TE values.

Resting‐State Functional Magnetic Resonance Imaging: The Impact of Regression Analysis

To investigate the impact of regression methods on resting-state functional magnetic resonance imaging (rsfMRI). During rsfMRI preprocessing, regression analysis is considered effective for reducing the interference of physiological noise on the signal time course. However, it is unclear whether the regression method benefits rsfMRI analysis. Twenty volunteers (10 men and 10 women; aged 23.4 ± 1.5 years) participated in the experiments. We used node analysis and functional connectivity mapping to assess the brain default mode network by using five combinations of regression methods. The results show that regressing the global mean plays a major role in the preprocessing steps. When a global regression method is applied, the values of functional connectivity are significantly lower (P ≤ .01) than those calculated without a global regression. This step increases inter-subject variation and produces anticorrelated brain areas. rsfMRI data processed using regression should be interpreted carefully. The significance of the anticorrelated brain areas produced by global signal removal is unclear.

SINGULAR SPECTRUM ANALYSIS AND ADAPTIVE FILTERING ENHANCE THE FUNCTIONAL CONNECTIVITY ANALYSIS OF RESTING STATE fMRI DATA

Sources of noise in resting-state fMRI experiments include instrumental and physiological noises, which need to be filtered before a functional connectivity analysis of brain regions is performed. These noisy components show autocorrelated and nonstationary properties that limit the efficacy of standard techniques (i.e. time filtering and general linear model). Herein we describe a novel approach based on the combination of singular spectrum analysis and adaptive filtering, which allows a greater noise reduction and yields better connectivity estimates between regions at rest, providing a new feasible procedure to analyze fMRI data.

Pupil diameter covaries with BOLD activity in human locus coeruleus

The locus coeruleus-noradrenergic (LC-NA) neuromodulatory system has been implicated in a broad array of cognitive processes, yet scope for investigating this system's function in humans is currently limited by an absence of reliable non-invasive measures of LC activity. Although pupil diameter has been employed as a proxy measure of LC activity in numerous studies, empirical evidence for a relationship between the two is lacking. In the present study, we sought to rigorously probe the relationship between pupil diameter and BOLD activity localized to the human LC. Simultaneous pupillometry and fMRI revealed a relationship between continuous pupil diameter and BOLD activity in a dorsal pontine cluster overlapping with the LC, as localized via neuromelanin-sensitive structural imaging and an LC atlas. This relationship was present both at rest and during performance of a two-stimulus oddball task, with and without spatial smoothing of the fMRI data, and survived retrospective image correction for physiological noise. Furthermore, the spatial extent of this pupil/LC relationship guided a volume-of-interest analysis in which we provide the first demonstration in humans of a fundamental characteristic of animal LC activity: phasic modulation by oddball stimulus relevance. Taken together, these findings highlight the potential for utilizing pupil diameter to achieve a more comprehensive understanding of the role of the LC-NA system in human cognition.

Myocardial arterial spin labeling perfusion imaging with improved sensitivity.

BackgroundMyocardial arterial spin labeling (ASL) is a noninvasive MRI based technique that is capable of measuring myocardial blood flow (MBF) in humans. It suffers from poor sensitivity to MBF due to high physiological noise (PN). This study aims to determine if the sensitivity of myocardial ASL to MBF can be improved by reducing image acquisition time, via parallel imaging.MethodsMyocardial ASL scans were performed in 7 healthy subjects at rest using flow-sensitive alternating inversion recovery (FAIR) tagging and balanced steady state free precession (SSFP) imaging. Sensitivity encoding (SENSE) with a reduction factor of 2 was used to shorten each image acquisition from roughly 300 ms per heartbeat to roughly 150 ms per heartbeat. A paired Student’s t-test was performed to compare measurements of myocardial blood flow (MBF) and physiological noise (PN) from the reference and accelerated methods.ResultsThe measured PN (mean ± standard deviation) was 0.20 ± 0.08 ml/g/min for the reference method and 0.08 ± 0.05 ml/g/min for the accelerated method, corresponding to a 60% reduction. PN measured from the accelerated method was found to be significantly lower than that of the reference method (p = 0.0059). There was no significant difference between MBF measured from the accelerated and reference ASL methods (p = 0.7297).ConclusionsIn this study, significant PN reduction was achieved by shortening the acquisition window using parallel imaging with no significant impact on the measured MBF. This indicates an improvement in sensitivity to MBF and may also enable the imaging of subjects with higher heart rates and imaging during systole.

Assessment of data acquisition parameters, and analysis techniques for noise reduction in spinal cord fMRI data

PurposeThe purpose of this work is to characterize the noise in spinal cord functional MRI, assess current methods aimed at reducing noise, and optimize imaging parameters. MethodsFunctional MRI data were acquired at multiple echo times and the contrast-to-noise ratio (CNR) was calculated. Independently, the repetition time was systematically varied with and without parallel imaging, to maximize BOLD sensitivity and minimize type I errors. Noise in the images was characterized by examining the frequency spectrum, and investigating whether autocorrelations exist. The efficacy of several physiological noise reduction methods in both null (no stimuli) and task (thermal pain paradigm) data was also assessed. Finally, our previous normalization methods were extended. ResultsThe echo time with the highest functional CNR at 3 Tesla is at roughly 75msec. Parallel imaging reduced the variance and the presence of autocorrelations, however the BOLD response in task data was more robust in data acquired without parallel imaging. Model-free based approaches further increased the detection of active voxels in the task data. Finally, inter-subject registration was improved. ConclusionsResults from this study provide a rigorous characterization of the properties of the noise and assessment of data acquisition and analysis methods for spinal cord and brainstem fMRI.

Enhanced phase regression with savitzky‐golay filtering for high‐resolution BOLD fMRI

Phase regression exploits the temporal evolution of phase in individual voxels to suppress blood oxygenation level dependent (BOLD) signal fluctuations caused by larger vessels and draining veins while preserving signal changes from microvascular effects. However, this process does not perform well when phase time series have low signal-to-noise ratios because of high levels of physiological noise. We demonstrate that Savitzky-Golay filters may be used to recover the underlying change in phase and completely restore the efficacy of phase regression. We do not make a priori assumptions regarding phase evolution and perform a data-driven exploration of parameter space to select the Savitzky-Golay filter parameters that minimize temporal variance in each voxel after phase regression. This approach is shown to work well on data acquired with single-shot and multi-shot pulse sequences, and should therefore be useful for both human and animal gradient-echo fMRI at high spatial resolutions at high fields. The ability to improve the spatial specificity of BOLD activation may be especially advantageous for clinical applications of fMRI that rely upon the accuracy of individual subject's activation maps to assist with presurgical planning and clinical decision-making. Enhanced phase regression with Savitzky-Golay filtering may also find other uses in analyses of resting state functional connectivity.

A methodology for fast assessments to the electrical activity of barrel fields in vivo: from population inputs to single unit outputs

Here we propose a methodology to analyze volumetric electrical activity of neuronal masses in the somatosensory barrel field of Wistar rats. The key elements of the proposed methodology are a three-dimensional microelectrode array, which was customized by our group to observe extracellular recordings from an extended area of the barrel field, and a novel method for the current source density analysis. By means of this methodology, we were able to localize single barrels from their event-related responses to single whisker deflection. It was also possible to assess the spatiotemporal dynamics of neuronal aggregates in several barrels at the same time with the resolution of single neurons. We used simulations to study the robustness of our methodology to unavoidable physiological noise and electrode configuration. We compared the accuracy to reconstruct neocortical current sources with that obtained with a previous method. This constitutes a type of electrophysiological microscopy with high spatial and temporal resolution, which could change the way we analyze the activity of cortical neurons in the future.

Characterization and reduction of cardiac- and respiratory-induced noise as a function of the sampling rate (TR) in fMRI

It has recently been shown that both high-frequency and low-frequency cardiac and respiratory noise sources exist throughout the entire brain and can cause significant signal changes in fMRI data. It is also known that the brainstem, basal forebrain and spinal cord areas are problematic for fMRI because of the magnitude of cardiac-induced pulsations at these locations. In this study, the physiological noise contributions in the lower brain areas (covering the brainstem and adjacent regions) are investigated and a novel method is presented for computing both low-frequency and high-frequency physiological regressors accurately for each subject. In particular, using a novel optimization algorithm that penalizes curvature (i.e. the second derivative) of the physiological hemodynamic response functions, the cardiac- and respiratory-related response functions are computed. The physiological noise variance is determined for each voxel and the frequency-aliasing property of the high-frequency cardiac waveform as a function of the repetition time (TR) is investigated. It is shown that for the brainstem and other brain areas associated with large pulsations of the cardiac rate, the temporal SNR associated with the low-frequency range of the BOLD response has maxima at subject-specific TRs. At these values, the high-frequency aliased cardiac rate can be eliminated by digital filtering without affecting the BOLD-related signal.