Purpose
To develop an improved k‐space reconstruction method using scan‐specific deep learning that is trained on autocalibration signal (ACS) data.
Theory
Robust artificial‐neural‐networks for ...k‐space interpolation (RAKI) reconstruction trains convolutional neural networks on ACS data. This enables nonlinear estimation of missing k‐space lines from acquired k‐space data with improved noise resilience, as opposed to conventional linear k‐space interpolation‐based methods, such as GRAPPA, which are based on linear convolutional kernels.
Methods
The training algorithm is implemented using a mean square error loss function over the target points in the ACS region, using a gradient descent algorithm. The neural network contains 3 layers of convolutional operators, with 2 of these including nonlinear activation functions. The noise performance and reconstruction quality of the RAKI method was compared with GRAPPA in phantom, as well as in neurological and cardiac in vivo data sets.
Results
Phantom imaging shows that the proposed RAKI method outperforms GRAPPA at high (≥4) acceleration rates, both visually and quantitatively. Quantitative cardiac imaging shows improved noise resilience at high acceleration rates (rate 4:23% and rate 5:48%) over GRAPPA. The same trend of improved noise resilience is also observed in high‐resolution brain imaging at high acceleration rates.
Conclusion
The RAKI method offers a training database‐free deep learning approach for MRI reconstruction, with the potential to improve many existing reconstruction approaches, and is compatible with conventional data acquisition protocols.
Purpose
To develop a strategy for training a physics‐guided MRI reconstruction neural network without a database of fully sampled data sets.
Methods
Self‐supervised learning via data undersampling ...(SSDU) for physics‐guided deep learning reconstruction partitions available measurements into two disjoint sets, one of which is used in the data consistency (DC) units in the unrolled network and the other is used to define the loss for training. The proposed training without fully sampled data is compared with fully supervised training with ground‐truth data, as well as conventional compressed‐sensing and parallel imaging methods using the publicly available fastMRI knee database. The same physics‐guided neural network is used for both proposed SSDU and supervised training. The SSDU training is also applied to prospectively two‐fold accelerated high‐resolution brain data sets at different acceleration rates, and compared with parallel imaging.
Results
Results on five different knee sequences at an acceleration rate of 4 shows that the proposed self‐supervised approach performs closely with supervised learning, while significantly outperforming conventional compressed‐sensing and parallel imaging, as characterized by quantitative metrics and a clinical reader study. The results on prospectively subsampled brain data sets, in which supervised learning cannot be used due to lack of ground‐truth reference, show that the proposed self‐supervised approach successfully performs reconstruction at high acceleration rates (4, 6, and 8). Image readings indicate improved visual reconstruction quality with the proposed approach compared with parallel imaging at acquisition acceleration.
Conclusion
The proposed SSDU approach allows training of physics‐guided deep learning MRI reconstruction without fully sampled data, while achieving comparable results with supervised deep learning MRI trained on fully sampled data.
Abstract
Functional magnetic resonance imaging (fMRI) has become an indispensable tool for investigating the human brain. However, the inherently poor signal-to-noise-ratio (SNR) of the fMRI ...measurement represents a major barrier to expanding its spatiotemporal scale as well as its utility and ultimate impact. Here we introduce a denoising technique that selectively suppresses the thermal noise contribution to the fMRI experiment. Using 7-Tesla, high-resolution human brain data, we demonstrate improvements in key metrics of functional mapping (temporal-SNR, the detection and reproducibility of stimulus-induced signal changes, and accuracy of functional maps) while leaving the amplitude of the stimulus-induced signal changes, spatial precision, and functional point-spread-function unaltered. We demonstrate that the method enables the acquisition of ultrahigh resolution (0.5 mm isotropic) functional maps but is also equally beneficial for a large variety of fMRI applications, including supra-millimeter resolution 3- and 7-Tesla data obtained over different cortical regions with different stimulation/task paradigms and acquisition strategies.
We evaluate residual aliasing among simultaneously excited and acquired slices in slice accelerated multiband (MB) echo planar imaging (EPI). No in-plane accelerations were used in order to maximize ...and evaluate achievable slice acceleration factors at 3T. We propose a novel leakage (L-) factor to quantify the effects of signal leakage between simultaneously acquired slices. With a standard 32-channel receiver coil at 3T, we demonstrate that slice acceleration factors of up to eight (MB=8) with blipped controlled aliasing in parallel imaging (CAIPI), in the absence of in-plane accelerations, can be used routinely with acceptable image quality and integrity for whole brain imaging. Spectral analyses of single-shot fMRI time series demonstrate that temporal fluctuations due to both neuronal and physiological sources were distinguishable and comparable up to slice-acceleration factors of nine (MB=9). The increased temporal efficiency could be employed to achieve, within a given acquisition period, higher spatial resolution, increased fMRI statistical power, multiple TEs, faster sampling of temporal events in a resting state fMRI time series, increased sampling of q-space in diffusion imaging, or more quiet time during a scan.
Display omitted
•High slice accelerations using multiband (MB) GRE-EPI with blipped CAIPI.•Acceptable MB factors up to 8 with a 32-channel receiver coil at 3T.•Neuronal and physiological sources are distinguishable at high MB factors.•Leakage (L-) factor evaluates residual aliasing among simultaneously acquired slices.•High temporal efficiency with MB-EPI benefits various applications.
Noninvasive human neuroimaging has yielded many discoveries about the brain. Numerous methodological advances have also occurred, though inertia has slowed their adoption. This paper presents an ...integrated approach to data acquisition, analysis and sharing that builds upon recent advances, particularly from the Human Connectome Project (HCP). The 'HCP-style' paradigm has seven core tenets: (i) collect multimodal imaging data from many subjects; (ii) acquire data at high spatial and temporal resolution; (iii) preprocess data to minimize distortions, blurring and temporal artifacts; (iv) represent data using the natural geometry of cortical and subcortical structures; (v) accurately align corresponding brain areas across subjects and studies; (vi) analyze data using neurobiologically accurate brain parcellations; and (vii) share published data via user-friendly databases. We illustrate the HCP-style paradigm using existing HCP data sets and provide guidance for future research. Widespread adoption of this paradigm should accelerate progress in understanding the brain in health and disease.
Functional magnetic resonance imaging (fMRI) studies that require high-resolution whole-brain coverage have long scan times that are primarily driven by the large number of thin slices acquired. ...Two-dimensional multiband echo-planar imaging (EPI) sequences accelerate the data acquisition along the slice direction and therefore represent an attractive approach to such studies by improving the temporal resolution without sacrificing spatial resolution. In this work, a 2D multiband EPI sequence was optimized for 1.5mm isotropic whole-brain acquisitions at 3T with 10 healthy volunteers imaged while performing simultaneous visual and motor tasks. The performance of the sequence was evaluated in terms of BOLD sensitivity and false-positive activation at multiband (MB) factors of 1, 2, 4, and 6, combined with in-plane GRAPPA acceleration of 2× (GRAPPA 2), and the two reconstruction approaches of Slice-GRAPPA and Split Slice-GRAPPA. Sensitivity results demonstrate significant gains in temporal signal-to-noise ratio (tSNR) and t-score statistics for MB 2, 4, and 6 compared to MB 1. The MB factor for optimal sensitivity varied depending on anatomical location and reconstruction method. When using Slice-GRAPPA reconstruction, evidence of false-positive activation due to signal leakage between simultaneously excited slices was seen in one instance, 35 instances, and 70 instances over the ten volunteers for the respective accelerations of MB 2×GRAPPA 2, MB 4×GRAPPA 2, and MB 6×GRAPPA 2. The use of Split Slice-GRAPPA reconstruction suppressed the prevalence of false positives significantly, to 1 instance, 5 instances, and 5 instances for the same respective acceleration factors. Imaging protocols using an acceleration factor of MB 2×GRAPPA 2 can be confidently used for high-resolution whole-brain imaging to improve BOLD sensitivity with very low probability for false-positive activation due to slice leakage. Imaging protocols using higher acceleration factors (MB 3 or MB 4×GRAPPA 2) can likely provide even greater gains in sensitivity but should be carefully optimized to minimize the possibility of false activations.
•MB factors 1, 2, 4, and 6 and two reconstructions were evaluated for fMRI performance.•MB accelerations 2, 4, and 6 improved BOLD sensitivity metrics over MB 1.•False-positive activation due to signal leakage was seen at high accelerations.•Use of Split Slice-GRAPPA reconstruction significantly reduces false positives.
■A hybrid linear and non-linear deep learning reconstruction approach for parallel imaging is proposed.■Improved image quality in highly accelerated MRI compared to conventional linear ...methods.■Improved interpretability compared to purely non-linear deep learning methods.■Improved preservation of temporal information in highly-accelerated SMS fMRI.
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.
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
ℓ
1
-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
ℓ
1
-wavelet CS can be fine-tuned to a level close to DL reconstruction for accelerated MRI. The optimized
ℓ
1
-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.
Echo planar imaging (EPI) is an MRI technique of particular value to neuroscience, with its use for virtually all functional MRI (fMRI) and diffusion imaging of fiber connections in the human brain. ...EPI generates a single 2D image in a fraction of a second; however, it requires 2-3 seconds to acquire multi-slice whole brain coverage for fMRI and even longer for diffusion imaging. Here we report on a large reduction in EPI whole brain scan time at 3 and 7 Tesla, without significantly sacrificing spatial resolution, and while gaining functional sensitivity. The multiplexed-EPI (M-EPI) pulse sequence combines two forms of multiplexing: temporal multiplexing (m) utilizing simultaneous echo refocused (SIR) EPI and spatial multiplexing (n) with multibanded RF pulses (MB) to achieve m×n images in an EPI echo train instead of the normal single image. This resulted in an unprecedented reduction in EPI scan time for whole brain fMRI performed at 3 Tesla, permitting TRs of 400 ms and 800 ms compared to a more conventional 2.5 sec TR, and 2-4 times reductions in scan time for HARDI imaging of neuronal fibertracks. The simultaneous SE refocusing of SIR imaging at 7 Tesla advantageously reduced SAR by using fewer RF refocusing pulses and by shifting fat signal out of the image plane so that fat suppression pulses were not required. In preliminary studies of resting state functional networks identified through independent component analysis, the 6-fold higher sampling rate increased the peak functional sensitivity by 60%. The novel M-EPI pulse sequence resulted in a significantly increased temporal resolution for whole brain fMRI, and as such, this new methodology can be used for studying non-stationarity in networks and generally for expanding and enriching the functional information.
Resting-state functional magnetic resonance imaging has become a powerful tool for the study of functional networks in the brain. Even "at rest," the brain's different functional networks ...spontaneously fluctuate in their activity level; each network's spatial extent can therefore be mapped by finding temporal correlations between its different subregions. Current correlation-based approaches measure the average functional connectivity between regions, but this average is less meaningful for regions that are part of multiple networks; one ideally wants a network model that explicitly allows overlap, for example, allowing a region's activity pattern to reflect one network's activity some of the time, and another network's activity at other times. However, even those approaches that do allow overlap have often maximized mutual spatial independence, which may be suboptimal if distinct networks have significant overlap. In this work, we identify functionally distinct networks by virtue of their temporal independence, taking advantage of the additional temporal richness available via improvements in functional magnetic resonance imaging sampling rate. We identify multiple "temporal functional modes," including several that subdivide the default-mode network (and the regions anticorrelated with it) into several functionally distinct, spatially overlapping, networks, each with its own pattern of correlations and anticorrelations. These functionally distinct modes of spontaneous brain activity are, in general, quite different from resting-state networks previously reported, and may have greater biological interpretability.