Whole-head rapid fMRI acquisition using echo-shifted magnetic resonance inverse imaging

Statistical power or more precise insights into neuro-temporal dynamics? Assessing the benefits of rapid temporal sampling in fMRI

Functional magnetic resonance imaging (fMRI), a non-invasive and widely used human neuroimaging method, is most known for its spatial precision. However, there is a growing interest in its temporal sensitivity. This is despite the temporal blurring of neuronal events by the blood oxygen level dependent (BOLD) signal, the peak of which lags neuronal firing by 4-6 seconds. Given this, the goal of this review is to answer a seemingly simple question - "What are the benefits of increased temporal sampling for fMRI?". To answer this, we have combined fMRI data collected at multiple temporal scales, from 323 to 1000 milliseconds, with a review of both historical and contemporary temporal literature. After a brief discussion of technological developments that have rekindled interest in temporal research, we next consider the potential statistical and methodological benefits. Most importantly, we explore how fast fMRI can uncover previously unobserved neuro-temporal dynamics - effects that are entirely missed when sampling at conventional 1 to 2 second rates. With the intrinsic link between space and time in fMRI, this temporal renaissance also delivers improvements in spatial precision. Far from producing only statistical gains, the array of benefits suggest that the continued temporal work is worth the effort.

15 Years MR-encephalography

Objective

This review article gives an account of the development of the MR-encephalography (MREG) method, which started as a mere ‘Gedankenexperiment’ in 2005 and gradually developed into a method for ultrafast measurement of physiological activities in the brain. After going through different approaches covering k-space with radial, rosette, and concentric shell trajectories we have settled on a stack-of-spiral trajectory, which allows full brain coverage with (nominal) 3 mm isotropic resolution in 100 ms. The very high acceleration factor is facilitated by the near-isotropic k-space coverage, which allows high acceleration in all three spatial dimensions.

Methods

The methodological section covers the basic sequence design as well as recent advances in image reconstruction including the targeted reconstruction, which allows real-time feedback applications, and—most recently—the time-domain principal component reconstruction (tPCR), which applies a principal component analysis of the acquired time domain data as a sparsifying transformation to improve reconstruction speed as well as quality.

Applications

Although the BOLD-response is rather slow, the high speed acquisition of MREG allows separation of BOLD-effects from cardiac and breathing related pulsatility. The increased sensitivity enables direct detection of the dynamic variability of resting state networks as well as localization of single interictal events in epilepsy patients. A separate and highly intriguing application is aimed at the investigation of the glymphatic system by assessment of the spatiotemporal patterns of cardiac and breathing related pulsatility.

Discussion

MREG has been developed to push the speed limits of fMRI. Compared to multiband-EPI this allows considerably faster acquisition at the cost of reduced image quality and spatial resolution.

Sampling Rate Effects on Resting State fMRI Metrics

Low image sampling rates used in resting state functional magnetic resonance imaging (rs-fMRI) may cause aliasing of the cardiorespiratory pulsations over the very low frequency (VLF) BOLD signal fluctuations which reflects to functional connectivity (FC). In this study, we examine the effect of sampling rate on currently used rs-fMRI FC metrics. Ultra-fast fMRI magnetic resonance encephalography (MREG) data, sampled with TR 0.1 s, was downsampled to different subsampled repetition times (sTR, range 0.3–3 s) for comparisons. Echo planar k-space sampling (TR 2.15 s) and interleaved slice collection schemes were also compared against the 3D single shot trajectory at 2.2 s sTR. The quantified connectivity metrics included stationary spatial, time, and frequency domains, as well as dynamic analyses. Time domain methods included analyses of seed-based functional connectivity, regional homogeneity (ReHo), coefficient of variation, and spatial domain group level probabilistic independent component analysis (ICA). In frequency domain analyses, we examined fractional and amplitude of low frequency fluctuations. Aliasing effects were spatially and spectrally analyzed by comparing VLF (0.01–0.1 Hz), respiratory (0.12–0.35 Hz) and cardiac power (0.9–1.3 Hz) FFT maps at different sTRs. Quasi-periodic pattern (QPP) of VLF events were analyzed for effects on dynamic FC methods. The results in conventional time and spatial domain analyses remained virtually unchanged by the different sampling rates. In frequency domain, the aliasing occurred mainly in higher sTR (1–2 s) where cardiac power aliases over respiratory power. The VLF power maps suffered minimally from increasing sTRs. Interleaved data reconstruction induced lower ReHo compared to 3D sampling (p < 0.001). Gradient recalled echo-planar imaging (EPI BOLD) data produced both better and worse metrics. In QPP analyses, the repeatability of the VLF pulse detection becomes linearly reduced with increasing sTR. In conclusion, the conventional resting state metrics (e.g., FC, ICA) were not markedly affected by different TRs (0.1–3 s). However, cardiorespiratory signals showed strongest aliasing in central brain regions in sTR 1–2 s. Pulsatile QPP and other dynamic analyses benefit linearly from short TR scanning.

Improving the signal detection accuracy of functional Magnetic Resonance Imaging

A major drawback of functional Magnetic Resonance Imaging (fMRI) concerns the lack of detection accuracy of the measured signal. Although this limitation stems in part from the neuro-vascular nature of the fMRI signal, it also reflects particular methodological decisions in the fMRI data analysis pathway. Here we show that the signal detection accuracy of fMRI is affected by the specific way in which whole-brain volumes are created from individually acquired brain slices, and by the method of statistically extracting signals from the sampled data. To address these limitations, we propose a new framework for fMRI data analysis. The new framework creates whole-brain volumes from individual brain slices that are all acquired at the same point in time relative to a presented stimulus. These whole-brain volumes contain minimal temporal distortions, and are available at a high temporal resolution. In addition, statistical signal extraction occurred on the basis of a non-standard time point-by-time point approach. We evaluated the detection accuracy of the extracted signal in the standard and new framework with simulated and real-world fMRI data. The new slice-based data-analytic framework yields greatly improved signal detection accuracy of fMRI signals.

Evaluation of SLIce Dithered Enhanced Resolution Simultaneous MultiSlice (SLIDER-SMS) for human fMRI

High isotropic resolution fMRI is challenging primarily due to long repetition times (TR) and insufficient SNR, especially at lower field strengths. Recently, Simultaneous Multi-Slice (SMS) imaging with blipped-CAIPI has substantially reduced scan time and improved SNR efficiency of fMRI. Similarly, super-resolution techniques utilizing sub- voxel spatial shifts in the slice direction have increased both resolution and SNR efficiency. Here we demonstrate the synergistic combination of SLIce Dithered Enhanced Resolution (SLIDER) and SMS for high-resolution, high-SNR whole brain fMRI in comparison to standard resolution fMRI data as well as high-resolution data. With SLIDER-SMS, high spatial frequency information is recovered (unaliased) even in absence of super-resolution deblurring algorithms. Additionally we find that BOLD CNR (as measured by t-value in a visual checkerboard paradigm) is improved by as much as 100% relative to traditionally acquired high- resolution data. Using this gain in CNR, we are able to obtain unprecedented nominally isotropic resolutions at 3T (0.66 mm) and 7T (0.45 mm).

Simultaneous multi-slice inverse imaging of the human brain

Ultrafast functional magnetic resonance imaging (fMRI) can measure blood oxygen level dependent (BOLD) signals with high sensitivity and specificity. Here we propose a novel method: simultaneous multi-slice inverse imaging (SMS-InI) — a combination of simultaneous multi-slice excitation, simultaneous echo refocusing (SER), blipped controlled aliasing in parallel imaging echo-planar imaging (EPI), and regularized image reconstruction. Using a 32-channel head coil array on a 3 T scanner, SMS-InI achieves nominal isotropic 5-mm spatial resolution and 10 Hz sampling rate at the whole-brain level. Compared with traditional inverse imaging, we found that SMS-InI has higher spatial resolution with lower signal leakage and higher time-domain signal-to-noise ratio with the optimized regularization parameter in the reconstruction. SMS-InI achieved higher effective resolution and higher detection power in detecting visual cortex activity than InI. SMS-InI also detected subcortical fMRI signals with the similar sensitivity and localization accuracy like EPI. The spatiotemporal resolution of SMS-InI was used to reveal that presenting visual stimuli with 0.2 s latency between left and right visual hemifield led to 0.2 s relative hemodynamic response latency between the left and right visual cortices. Together, these results indicate that SMS-InI is a useful tool in measuring cortical and subcortical hemodynamic responses with high spatiotemporal resolution.

Focused Ultrasound Help Realize High Spatiotemporal Brain Imaging?-A Concept on Acousto-Electrophysiological Neuroimaging

We present a concept for integrating focused ultrasound (FUS) with electrophysiological neuroimaging, in order to achieve high spatiotemporal resolution brain imaging. This approach, which we are tentatively calling acousto-electrophysiological neuroimaging, leverages on the spatial focality and noninvasiveness of FUS and may potentially lead to a noninvasive human brain imaging modality with high resolution in both space and time domains. By the use of modulated FUS, spatial selectivity can be accomplished for high-resolution electrophysiological neuroimaging. Frequency shifting in resulting magnetic signals (using modulated FUS) may potentially open the door for a room temperature magnetoencephalography device.

Using precise word timing information improves decoding accuracy in a multiband-accelerated multimodal reading experiment

The blood-oxygen-level-dependent (BOLD) signal measured in functional magnetic resonance imaging (fMRI) experiments is generally regarded as sluggish and poorly suited for probing neural function at the rapid timescales involved in sentence comprehension. However, recent studies have shown the value of acquiring data with very short repetition times (TRs), not merely in terms of improvements in contrast to noise ratio (CNR) through averaging, but also in terms of additional fine-grained temporal information. Using multiband-accelerated fMRI, we achieved whole-brain scans at 3-mm resolution with a TR of just 500 ms at both 3T and 7T field strengths. By taking advantage of word timing information, we found that word decoding accuracy across two separate sets of scan sessions improved significantly, with better overall performance at 7T than at 3T. The effect of TR was also investigated; we found that substantial word timing information can be extracted using fast TRs, with diminishing benefits beyond TRs of 1000 ms.

7T-fMRI: Faster temporal resolution yields optimal BOLD sensitivity for functional network imaging specifically at high spatial resolution

Recent developments in accelerated imaging methods allow faster acquisition of high spatial resolution images. This could improve the applications of functional magnetic resonance imaging at 7 Tesla (7T-fMRI), such as neurosurgical planning and Brain Computer Interfaces (BCIs). However, increasing the spatial and temporal resolution will both lead to signal-to-noise ratio (SNR) losses due to decreased net magnetization per voxel and T₁-relaxation effect, respectively. This could potentially offset the SNR efficiency gains made with increasing temporal resolution. We investigated the effects of varying spatial and temporal resolution on fMRI sensitivity measures and their implications on fMRI-based BCI simulations. We compared temporal signal-to-noise ratio (tSNR), observed percent signal change (%∆S), volumes of significant activation, Z-scores and decoding performance of linear classifiers commonly used in BCIs across a range of spatial and temporal resolution images acquired during an ankle-tapping task. Our results revealed an average increase of 22% in %∆S (p=0.006) and 9% in decoding performance (p=0.015) with temporal resolution only at the highest spatial resolution of 1.5×1.5×1.5mm³, despite a 29% decrease in tSNR (p<0.001) and plateaued Z-scores. Further, the volume of significant activation was indifferent (p>0.05) across spatial resolution specifically at the highest temporal resolution of 500ms. These results demonstrate that the overall BOLD sensitivity can be increased significantly with temporal resolution, granted an adequately high spatial resolution with minimal physiological noise level. This shows the feasibility of diffuse motor-network imaging at high spatial and temporal resolution with robust BOLD sensitivity with 7T-fMRI. Importantly, we show that this sensitivity improvement could be extended to an fMRI application such as BCIs.

Increased fMRI Sensitivity at Equal Data Burden Using Averaged Shifted Echo Acquisition

There is growing evidence as to the benefits of collecting BOLD fMRI data with increased sampling rates. However, many of the newly developed acquisition techniques developed to collect BOLD data with ultra-short TRs require hardware, software, and non-standard analytic pipelines that may not be accessible to all researchers. We propose to incorporate the method of shifted echo into a standard multi-slice, gradient echo EPI sequence to achieve a higher sampling rate with a TR of <1 s with acceptable spatial resolution. We further propose to incorporate temporal averaging of consecutively acquired EPI volumes to both ameliorate the reduced temporal signal-to-noise inherent in ultra-fast EPI sequences and reduce the data burden. BOLD data were collected from 11 healthy subjects performing a simple, event-related visual-motor task with four different EPI sequences: (1) reference EPI sequence with TR = 1440 ms, (2) shifted echo EPI sequence with TR = 700 ms, (3) shifted echo EPI sequence with every two consecutively acquired EPI volumes averaged and effective TR = 1400 ms, and (4) shifted echo EPI sequence with every four consecutively acquired EPI volumes averaged and effective TR = 2800 ms. Both the temporally averaged sequences exhibited increased temporal signal-to-noise over the shifted echo EPI sequence. The shifted echo sequence with every two EPI volumes averaged also had significantly increased BOLD signal change compared with the other three sequences, while the shifted echo sequence with every four EPI volumes averaged had significantly decreased BOLD signal change compared with the other three sequences. The results indicated that incorporating the method of shifted echo into a standard multi-slice EPI sequence is a viable method for achieving increased sampling rate for collecting event-related BOLD data. Further, consecutively averaging every two consecutively acquired EPI volumes significantly increased the measured BOLD signal change and the subsequently calculated activation map statistics.

An exploration of task based fMRI in neonates using echo-shifting to allow acquisition at longer TE without loss of temporal efficiency

Optimal contrast to noise ratio of the BOLD signal in neonatal and foetal fMRI has been hard to achieve because of the much longer T2(⁎) values in developing brain tissue in comparison to those in the mature adult brain. The conventional approach of optimizing fMRI sequences would suggest matching the echo time (TE) and the T2(⁎) of the neonatal and foetal brain. However, the use of a long echo time would typically increase the minimum repetition time (TR) resulting in inefficient sampling. Here we apply the concept of echo shifting to task based neonatal fMRI in order to achieve an improved contrast to noise ratio and efficient data sampling at the same time. Echo shifted EPI (es-EPI) is a modification of a standard 2D-EPI sequence which enables echo times longer than the time between consecutive excitations (TE>TS=TRNS, where NS is the number of acquired slices and TS the inter-slice repetition time). The proposed method was tested on neonatal subjects using a passive sensori-motor task paradigm. Dual echo EPI datasets with an identical readout structure to es-EPI were also acquired and used as control data to assess BOLD activation. From the results of the latter analysis, an average increase of 78±41% in contrast to noise ratio was observable when comparing late to short echoes. Furthermore, es-EPI allowed the acquisition of data with an identical contrast to the late echo, but more efficiently since a higher number of slices could be acquired in the same amount of time.

Advantages and disadvantages of a fast fMRI sequence in the context of EEG-fMRI investigation of epilepsy patients: A realistic simulation study

EEG-fMRI is an established technique to allow mapping BOLD changes in response to interictal discharges recorded in the EEG of epilepsy patients. Traditional fMRI experiments rely on an echo planar imaging (EPI) sequence with a time resolution given by its time-to-repetition (TR) of ~2 s. Recently, multiple fast fMRI sequences have been developed to get around the limited temporal resolution of the EPI sequence, and achieved a TR in the 100 ms range or lower. One such sequence is called Magnetic Resonance EncephaloGraphy (MREG). Its high temporal resolution should offer increased detection sensitivity and statistical power in the context of epilepsy studies and in fMRI experiments in general. The aim of this work was to investigate the advantages and disadvantages offered by MREG. This was done by superimposing artificial event-related BOLD responses on EPI and MREG background signals, from 5 epileptic patients, that were free of epileptic discharges (spikes) on simultaneously recorded EEG. These functional datasets simulated different spiking rates and hemodynamic response amplitudes, and were analyzed with the commonly used General Linear Model (GLM) with the canonical hemodynamic response function (HRF) as a fixed model of the response. Robustness to violation of the assumptions of the GLM was additionally assessed with similar simulations using variable spike-to-spike response amplitudes and 8 non-canonical HRFs. Consistent with previous work, MREG yields higher maximum statistical t-values than EPI, but our simulations showed these statistics to be inflated, as the false positive rate at a standard threshold was high. At thresholds set to appropriately control specificity, EPI showed better true positive rate and larger cluster size than MREG. However, the lack of an appropriate calibration of the amplitude of the responses across the sequences precludes definitive judgment on their relative sensitivity. In addition, we show that a mismatch between the assumed and actual HRF impairs more MREG detection performance, but that EPI is more affected by non-modeled spike-to-spike variations of response amplitude. Filtering-out physiological noise, which is not aliased at the fast sampling rate of MREG, and the modeling of temporal autocorrelation are advantageous in increasing the detection power of MREG. This simulation study 1) warrants care when interpreting statistical t-values from fast fMRI sequences, 2) proposes thresholds for valid inferences and processing methods for maximal sensitivities, and 3) demonstrates the relative robustness/susceptibility of MREG and EPI to violation of the GLM's assumptions.

Electrocorticography Reveals Enhanced Visual Cortex Responses to Visual Speech

Human speech contains both auditory and visual components, processed by their respective sensory cortices. We test a simple model in which task-relevant speech information is enhanced during cortical processing. Visual speech is most important when the auditory component is uninformative. Therefore, the model predicts that visual cortex responses should be enhanced to visual-only (V) speech compared with audiovisual (AV) speech. We recorded neuronal activity as patients perceived auditory-only (A), V, and AV speech. Visual cortex showed strong increases in high-gamma band power and strong decreases in alpha-band power to V and AV speech. Consistent with the model prediction, gamma-band increases and alpha-band decreases were stronger for V speech. The model predicts that the uninformative nature of the auditory component (not simply its absence) is the critical factor, a prediction we tested in a second experiment in which visual speech was paired with auditory white noise. As predicted, visual speech with auditory noise showed enhanced visual cortex responses relative to AV speech. An examination of the anatomical locus of the effects showed that all visual areas, including primary visual cortex, showed enhanced responses. Visual cortex responses to speech are enhanced under circumstances when visual information is most important for comprehension.

Improving the spatial resolution of magnetic resonance inverse imaging via the blipped-CAIPI acquisition scheme

Using simultaneous acquisition from multiple channels of a radio-frequency (RF) coil array, magnetic resonance inverse imaging (InI) achieves functional MRI acquisitions at a rate of 100ms per whole-brain volume. InI accelerates the scan by leaving out partition encoding steps and reconstructs images by solving under-determined inverse problems using RF coil sensitivity information. Hence, the correlated spatial information available in the coil array causes spatial blurring in the InI reconstruction. Here, we propose a method that employs gradient blips in the partition encoding direction during the acquisition to provide extra spatial encoding in order to better differentiate signals from different partitions. According to our simulations, this blipped-InI (bInI) method can increase the average spatial resolution by 15.1% (1.3mm) across the whole brain and from 32.6% (4.2mm) in subcortical regions, as compared to the InI method. In a visual fMRI experiment, we demonstrate that, compared to InI, the spatial distribution of bInI BOLD response is more consistent with that of a conventional echo-planar imaging (EPI) at the level of individual subjects. With the improved spatial resolution, especially in subcortical regions, bInI can be a useful fMRI tool for obtaining high spatiotemporal information for clinical and cognitive neuroscience studies.

Lag-based effective connectivity applied to fMRI: a simulation study highlighting dependence on experimental parameters and formulation

A vast repertoire of methods is currently available to study effective brain connectivity based on neuroimaging data, among which lag-based measures can be distinguished. Although several studies have previously assessed the performance of such measures, their validity in different conditions remains unclear. In the current study, several lag-based effective connectivity measures are tested and benchmarked using simulated fMRI data, conceived to reflect a broad range of different situations with practical interest. The main goal is two-fold: 1) to provide a thorough overview of lag-based effective connectivity measures, and 2) to assess their performance in specific experimental conditions, thereby providing guidance for future effective connectivity studies involving fMRI. We focus on well-known lag-based measures, cover existing improvements and alternative formulations in some cases: Granger causality (GC), Geweke's Granger causality (GGC), directed transfer function (DTF), partial directed coherence (PDC), phase slope index (PSI), and transfer entropy (TE). Benchmarking consists in identifying causal relations in local field potential (LFP) networks that have their output convolved with a canonical hemodynamic response function (HRF) with varying node number, topology, coupling strength, neuronal delay, repetition time (TR), signal-to-noise ratio (SNR) and HRF variability. In a first set of simulations, we cover all possible combinations of discretized values of the previous variables, for networks with 2 and 3 nodes, and find that the measure with best performance (time-domain Granger Causality) is able to detect neuronal delays of a few hundreds of milliseconds with TRs between 0.25 and 2s and neuronal delays below 100ms for TRs that are also below 100ms, with more than 80% accuracy in realistic conditions. For networks with more than 3 nodes, we find that the number of nodes and the density of causal links degrade sensitivity, especially if the number of observations does not compensate for the increase in nodes, and that clustered networks can be more easily identified. In conclusion, this study argues in favor of the applicability of lag-based measures in the context of fMRI, provided that a stringent set of experimental specifications is met and that the chosen measure is applied with full knowledge of its limitations and specific constraints.

Grand challenges in mapping the human brain: NSF workshop report

This report summarizes the outcomes of the NSF Workshop on Mapping and Engineering the Brain, held at Arlington, VA, during August 13-14, 2013. Three grand challenges were identified, including high spatiotemporal resolution neuroimaging, perturbation-based neuroimaging, and neuroimaging in naturalistic environments. It was highlighted that each grand challenge requires groundbreaking discoveries, enabling technologies, appropriate knowledge transfer, and multi- and transdisciplinary education and training for success.

Representation of frequency-modulated sounds in the human brain

Frequency-modulation is a ubiquitous sound feature present in communicative sounds of various animal species and humans. Functional imaging of the human auditory system has seen remarkable advances in the last two decades and studies pertaining to frequency-modulation have centered around two major questions: a) are there dedicated feature-detectors encoding frequency-modulation in the brain and b) is there concurrent representation with amplitude-modulation, another temporal sound feature? In this review, we first describe how these two questions are motivated by psychophysical studies and neurophysiology in animal models. We then review how human non-invasive neuroimaging studies have furthered our understanding of the representation of frequency-modulated sounds in the brain. Finally, we conclude with some suggestions on how human neuroimaging could be used in future studies to address currently still open questions on this fundamental sound feature. This article is part of a Special Issue entitled Human Auditory Neuroimaging.