Revisiting [Formula: see text]-wavelet compressed-sensing MRI in the era of deep learning.

Following their success in numerous imaging and computer vision applications, deep-learning (DL) techniques have emerged as one of the most prominent strategies for accelerated MRI reconstruction. These methods have been shown to outperform conventional regularized methods based on compressed sensing (CS). However, in most comparisons, CS is implemented with two or three hand-tuned parameters, while DL methods enjoy a plethora of advanced data science tools. In this work, we revisit [Formula: see text]-wavelet CS reconstruction using these modern tools. Using ideas such as algorithm unrolling and advanced optimization methods over large databases that DL algorithms utilize, along with conventional insights from wavelet representations and CS theory, we show that [Formula: see text]-wavelet CS can be fine-tuned to a level close to DL reconstruction for accelerated MRI. The optimized [Formula: see text]-wavelet CS method uses only 128 parameters compared to >500,000 for DL, employs a convex reconstruction at inference time, and performs within <1% of a DL approach that has been used in multiple studies in terms of quantitative quality metrics.

Unsupervised deep learning with convolutional neural networks for static parallel transmit design: A retrospective study.

PURPOSE: To mitigate B 1 + $$ {B}_1^{+} $$ inhomogeneity at 7T for multi-channel transmit arrays using unsupervised deep learning with convolutional neural networks (CNNs). METHODS: Deep learning parallel transmit (pTx) pulse design has received attention, but such methods have relied on supervised training and did not use CNNs for multi-channel B 1 + $$ {B}_1^{+} $$ maps. In this work, we introduce an alternative approach that facilitates the use of CNNs with multi-channel B 1 + $$ {B}_1^{+} $$ maps while performing unsupervised training. The multi-channel B 1 + $$ {B}_1^{+} $$ maps are concatenated along the spatial dimension to enable shift-equivariant processing amenable to CNNs. Training is performed in an unsupervised manner using a physics-driven loss function that minimizes the discrepancy of the Bloch simulation with the target magnetization, which eliminates the calculation of reference transmit RF weights. The training database comprises 3824 2D sagittal, multi-channel B 1 + $$ {B}_1^{+} $$ maps of the healthy human brain from 143 subjects. B 1 + $$ {B}_1^{+} $$ data were acquired at 7T using an 8Tx/32Rx head coil. The proposed method is compared to the unregularized magnitude least-squares (MLS) solution for the target magnetization in static pTx design. RESULTS: The proposed method outperformed the unregularized MLS solution for RMS error and coefficient-of-variation and had comparable energy consumption. Additionally, the proposed method did not show local phase singularities leading to distinct holes in the resulting magnetization unlike the unregularized MLS solution. CONCLUSION: Proposed unsupervised deep learning with CNNs performs better than unregularized MLS in static pTx for speed and robustness.

Performance of receive head arrays versus ultimate intrinsic SNR at 7 T and 10.5 T.

PURPOSE: We examined magnetic field dependent SNR gains and ability to capture them with multichannel receive arrays for human head imaging in going from 7 T, the most commonly used ultrahigh magnetic field (UHF) platform at the present, to 10.5 T, which represents the emerging new frontier of >10 T in UHFs. METHODS: Electromagnetic (EM) models of 31-channel and 63-channel multichannel arrays built for 10.5 T were developed for 10.5 T and 7 T simulations. A 7 T version of the 63-channel array with an identical coil layout was also built. Array performance was evaluated in the EM model using a phantom mimicking the size and electrical properties of the human head and a digital human head model. Experimental data was obtained at 7 T and 10.5 T with the 63-channel array. Ultimate intrinsic SNR (uiSNR) was calculated for the two field strengths using a voxelized cloud of dipoles enclosing the phantom or the digital human head model as a reference to assess the performance of the two arrays and field depended SNR gains. RESULTS: uiSNR calculations in both the phantom and the digital human head model demonstrated SNR gains at 10.5 T relative to 7 T of 2.6 centrally, ˜2 at the location corresponding to the edge of the brain, ˜1.4 at the periphery. The EM models demonstrated that, centrally, both arrays captured ˜90% of the uiSNR at 7 T, but only ˜65% at 10.5 T, leading only to ˜2-fold gain in array SNR in going from 7 to 10.5 T. This trend was also observed experimentally with the 63-channel array capturing a larger fraction of the uiSNR at 7 T compared to 10.5 T, although the percentage of uiSNR captured were slightly lower at both field strengths compared to EM simulation results. CONCLUSIONS: Major uiSNR gains are predicted for human head imaging in going from 7 T to 10.5 T, ranging from ˜2-fold at locations corresponding to the edge of the brain to 2.6-fold at the center, corresponding to approximately quadratic increase with the magnetic field. Realistic 31- and 63-channel receive arrays, however, approach the central uiSNR at 7 T, but fail to do so at 10.5 T, suggesting that more coils and/or different type of coils will be needed at 10.5 T and higher magnetic fields.

The possible influence of third-order shim coils on gradient-magnet interactions: an inter-field and inter-site study.

OBJECTIVE: To assess the possible influence of third-order shim coils on the behavior of the gradient field and in gradient-magnet interactions at 7 T and above. MATERIALS AND METHODS: Gradient impulse response function measurements were performed at 5 sites spanning field strengths from 7 to 11.7 T, all of them sharing the same exact whole-body gradient coil design. Mechanical fixation and boundary conditions of the gradient coil were altered in several ways at one site to study the impact of mechanical coupling with the magnet on the field perturbations. Vibrations, power deposition in the He bath, and field dynamics were characterized at 11.7 T with the third-order shim coils connected and disconnected inside the Faraday cage. RESULTS: For the same whole-body gradient coil design, all measurements differed greatly based on the third-order shim coil configuration (connected or not). Vibrations and gradient transfer function peaks could be affected by a factor of 2 or more, depending on the resonances. Disconnecting the third-order shim coils at 11.7 T also suppressed almost completely power deposition peaks at some frequencies. DISCUSSION: Third-order shim coil configurations can have major impact in gradient-magnet interactions with consequences on potential hardware damage, magnet heating, and image quality going beyond EPI acquisitions.

Evaluating increases in sensitivity from NORDIC for diverse fMRI acquisition strategies.

As the neuroimaging field moves towards detecting smaller effects at higher spatial resolutions, and faster sampling rates, there is increased attention given to the deleterious contribution of unstructured, thermal noise. Here, we critically evaluate the performance of a recently developed reconstruction method, termed NORDIC, for suppressing thermal noise using datasets acquired with various field strengths, voxel sizes, sampling rates, and task designs. Following minimal preprocessing, statistical activation (t-values) of NORDIC processed data was compared to the results obtained with alternative denoising methods. Additionally, we examined the consistency of the estimates of task responses at the single-voxel, single run level, using a finite impulse response (FIR) model. To examine the potential impact on effective image resolution, the overall smoothness of the data processed with different methods was estimated. Finally, to determine if NORDIC alters or removes temporal information important for modeling responses, we employed an exhaustive leave-p-out cross validation approach, using FIR task responses to predict held out timeseries, quantified using R2. After NORDIC, the t-values are increased, an improvement comparable to what could be achieved by 1.5 voxels smoothing, and task events are clearly visible and have less cross-run error. These advantages are achieved with smoothness estimates increasing by less than 4%, while 1.5 voxel smoothing is associated with increases of over 140%. Cross-validated R2s based on the FIR models show that NORDIC is not measurably distorting the temporal structure of the data under this approach and is the best predictor of non-denoised time courses. The results demonstrate that analyzing 1 run of data after NORDIC produces results equivalent to using 2 to 3 original runs and that NORDIC performs equally well across a diverse array of functional imaging protocols. Significance Statement: For functional neuroimaging, the increasing availability of higher field strengths and ever higher spatiotemporal resolutions has led to concomitant increase in concerns about the deleterious effects of thermal noise. Historically this noise source was suppressed using methods that reduce spatial precision such as image blurring or averaging over a large number of trials or sessions, which necessitates large data collection efforts. Here, we critically evaluate the performance of a recently developed reconstruction method, termed NORDIC, which suppresses thermal noise. Across datasets varying in field strength, voxel sizes, sampling rates, and task designs, NORDIC produces substantial gains in data quality. Both conventional t-statistics derived from general linear models and coefficients of determination for predicting unseen data are improved. These gains match or even exceed those associated with 1 voxel Full Width Half Max image smoothing, however, even such small amounts of smoothing are associated with a 52% reduction in estimates of spatial precision, whereas the measurable difference in spatial precision is less than 4% following NORDIC.

A temporal decomposition method for identifying venous effects in task-based fMRI.

The spatial resolution of functional magnetic resonance imaging (fMRI) is fundamentally limited by effects from large draining veins. Here we describe an analysis method that provides data-driven estimates of these effects in task-based fMRI. The method involves fitting a one-dimensional manifold that characterizes variation in response timecourses observed in a given dataset, and then using identified early and late timecourses as basis functions for decomposing responses into components related to the microvasculature (capillaries and small venules) and the macrovasculature (large veins), respectively. We show the removal of late components substantially reduces the superficial cortical depth bias of fMRI responses and helps eliminate artifacts in cortical activity maps. This method provides insight into the origins of the fMRI signal and can be used to improve the spatial accuracy of fMRI.

Reconstruction for 7T high-resolution whole-brain diffusion MRI using two-stage N/2 ghost correction and L1-SPIRiT without single-band reference.

PURPOSE: To combine a new two-stage N/2 ghost correction and an adapted L1-SPIRiT method for reconstruction of 7T highly accelerated whole-brain diffusion MRI (dMRI) using only autocalibration scans (ACS) without the need of additional single-band reference (SBref) scans. METHODS: The proposed ghost correction consisted of a 3-line reference approach in stage 1 and the reference-free entropy method in stage 2. The adapted L1-SPIRiT method was formulated within the 3D k-space framework. Its efficacy was examined by acquiring two dMRI data sets at 1.05-mm isotropic resolutions with a total acceleration of 6 or 9 (i.e., 2-fold or 3-fold slice and 3-fold in-plane acceleration). Diffusion analysis was performed to derive DTI metrics and estimate fiber orientation distribution functions (fODFs). The results were compared with those of 3D k-space GRAPPA using only ACS, all in reference to 3D k-space GRAPPA using both ACS and SBref (serving as a reference). RESULTS: The proposed ghost correction eliminated artifacts more robustly than conventional approaches. Our adapted L1-SPIRiT method outperformed 3D k-space GRAPPA when using only ACS, improving image quality to what was achievable with 3D k-space GRAPPA using both ACS and SBref scans. The improvement in image quality further resulted in an improvement in estimation performances for DTI and fODFs. CONCLUSION: The combination of our new ghost correction and adapted L1-SPIRiT method can reliably reconstruct 7T highly accelerated whole-brain dMRI without the need of SBref scans, increasing acquisition efficiency and reducing motion sensitivity.

Residual RAKI: A hybrid linear and non-linear approach for scan-specific k-space deep learning.

Parallel imaging is the most clinically used acceleration technique for magnetic resonance imaging (MRI) in part due to its easy inclusion into routine acquisitions. In k-space based parallel imaging reconstruction, sub-sampled k-space data are interpolated using linear convolutions. At high acceleration rates these methods have inherent noise amplification and reduced image quality. On the other hand, non-linear deep learning methods provide improved image quality at high acceleration, but the availability of training databases for different scans, as well as their interpretability hinder their adaptation. In this work, we present an extension of Robust Artificial-neural-networks for k-space Interpolation (RAKI), called residual-RAKI (rRAKI), which achieves scan-specific machine learning reconstruction using a hybrid linear and non-linear methodology. In rRAKI, non-linear CNNs are trained jointly with a linear convolution implemented via a skip connection. In effect, the linear part provides a baseline reconstruction, while the non-linear CNN that runs in parallel provides further reduction of artifacts and noise arising from the linear part. The explicit split between the linear and non-linear aspects of the reconstruction also help improve interpretability compared to purely non-linear methods. Experiments were conducted on the publicly available fastMRI datasets, as well as high-resolution anatomical imaging, comparing GRAPPA and its variants, compressed sensing, RAKI, Scan Specific Artifact Reduction in K-space (SPARK) and the proposed rRAKI. Additionally, highly-accelerated simultaneous multi-slice (SMS) functional MRI reconstructions were also performed, where the proposed rRAKI was compred to Read-out SENSE-GRAPPA and RAKI. Our results show that the proposed rRAKI method substantially improves the image quality compared to conventional parallel imaging, and offers sharper images compared to SPARK and ℓ1-SPIRiT. Furthermore, rRAKI shows improved preservation of time-varying dynamics compared to both parallel imaging and RAKI in highly-accelerated SMS fMRI.

Ultra-high field (10.5T) diffusion-weighted MRI of the macaque brain.

Diffu0sion-weighted magnetic resonance imaging (dMRI) is a non-invasive imaging technique that provides information about the barriers to the diffusion of water molecules in tissue. In the brain, this information can be used in several important ways, including to examine tissue abnormalities associated with brain disorders and to infer anatomical connectivity and the organization of white matter bundles through the use of tractography algorithms. However, dMRI also presents certain challenges. For example, historically, the biological validation of tractography models has shown only moderate correlations with anatomical connectivity as determined through invasive tract-tracing studies. Some of the factors contributing to such issues are low spatial resolution, low signal-to-noise ratios, and long scan times required for high-quality data, along with modeling challenges like complex fiber crossing patterns. Leveraging the capabilities provided by an ultra-high field scanner combined with denoising, we have acquired whole-brain, 0.58 mm isotropic resolution dMRI with a 2D-single shot echo planar imaging sequence on a 10.5 Tesla scanner in anesthetized macaques. These data produced high-quality tractograms and maps of scalar diffusion metrics in white matter. This work demonstrates the feasibility and motivation for in-vivo dMRI studies seeking to benefit from ultra-high fields.

Cortical layer-specific differences in stimulus selectivity revealed with high-field fMRI and single-vessel resolution optical imaging of the primary visual cortex.

The mammalian neocortex exhibits a stereotypical laminar organization, with feedforward inputs arriving primarily into layer 4, local computations shaping response selectivity in layers 2/3, and outputs to other brain areas emanating via layers 2/3, 5 and 6. It cannot be assumed a priori that these signatures of laminar differences in neuronal circuitry are reflected in hemodynamic signals that form the basis of functional magnetic resonance imaging (fMRI). Indeed, optical imaging of single-vessel functional responses has highlighted the potential limits of using vascular signals as surrogates for mapping the selectivity of neural responses. Therefore, before fMRI can be employed as an effective tool for studying critical aspects of laminar processing, validation with single-vessel resolution is needed. The primary visual cortex (V1) in cats, with its precise neuronal functional micro-architecture, offers an ideal model system to examine laminar differences in stimulus selectivity across imaging modalities. Here we used cerebral blood volume weighted (wCBV) fMRI to examine if layer-specific orientation-selective responses could be detected in cat V1. We found orientation preference maps organized tangential to the cortical surface that typically extended across depth in a columnar fashion. We then examined arterial dilation and blood velocity responses to identical visual stimuli by using two- and three- photon optical imaging at single-vessel resolution-which provides a measure of the hemodynamic signals with the highest spatial resolution. Both fMRI and optical imaging revealed a consistent laminar response pattern in which orientation selectivity in cortical layer 4 was significantly lower compared to layer 2/3. This systematic change in selectivity across cortical layers has a clear underpinning in neural circuitry, particularly when comparing layer 4 to other cortical layers.