Adapting model-based deep learning to multiple acquisition conditions: Ada-MoDL.

PURPOSE: The aim of this work is to introduce a single model-based deep network that can provide high-quality reconstructions from undersampled parallel MRI data acquired with multiple sequences, acquisition settings, and field strengths. METHODS: A single unrolled architecture, which offers good reconstructions for multiple acquisition settings, is introduced. The proposed scheme adapts the model to each setting by scaling the convolutional neural network (CNN) features and the regularization parameter with appropriate weights. The scaling weights and regularization parameter are derived using a multilayer perceptron model from conditional vectors, which represents the specific acquisition setting. The perceptron parameters and the CNN weights are jointly trained using data from multiple acquisition settings, including differences in field strengths, acceleration, and contrasts. The conditional network is validated using datasets acquired with different acquisition settings. RESULTS: The comparison of the adaptive framework, which trains a single model using the data from all the settings, shows that it can offer consistently improved performance for each acquisition condition. The comparison of the proposed scheme with networks that are trained independently for each acquisition setting shows that it requires less training data per acquisition setting to offer good performance. CONCLUSION: The Ada-MoDL framework enables the use of a single model-based unrolled network for multiple acquisition settings. In addition to eliminating the need to train and store multiple networks for different acquisition settings, this approach reduces the training data needed for each acquisition setting.

Split-slice training and hyperparameter tuning of RAKI networks for simultaneous multi-slice reconstruction.

PURPOSE: Simultaneous multi-slice acquisitions are essential for modern neuroimaging research, enabling high temporal resolution functional and high-resolution q-space sampling diffusion acquisitions. Recently, deep learning reconstruction techniques have been introduced for unaliasing these accelerated acquisitions, and robust artificial-neural-networks for k-space interpolation (RAKI) have shown promising capabilities. This study systematically examines the impacts of hyperparameter selections for RAKI networks, and introduces a novel technique for training data generation which is analogous to the split-slice formalism used in slice-GRAPPA. METHODS: RAKI networks were developed with variable hyperparameters and with and without split-slice training data generation. Each network was trained and applied to five different datasets including acquisitions harmonized with Human Connectome Project lifespan protocol. Unaliasing performance was assessed through L1 errors computed between unaliased and calibration frequency-space data. RESULTS: Split-slice training significantly improved network performance in nearly all hyperparameter configurations. Best unaliasing results were achieved with three layer RAKI networks using at least 64 convolutional filters with receptive fields of 7 voxels, 128 single-voxel filters in the penultimate RAKI layer, batch normalization, and no training dropout with the split-slice augmented training dataset. Networks trained without the split-slice technique showed symptoms of network over-fitting. CONCLUSIONS: Split-slice training for simultaneous multi-slice RAKI networks positively impacts network performance. Hyperparameter tuning of such reconstruction networks can lead to further improvements in unaliasing performance.

Multispectral diffusion-weighted MRI of the instrumented cervical spinal cord: a preliminary study of 5 cases.

PURPOSE: Diffusion-weighted imaging has undergone substantial investigation as a potential tool for advanced assessment of spinal cord health. Unfortunately, commonly encountered surgically implanted spinal hardware has historically disrupted these studies. This preliminary investigation applies the recently developed multispectral diffusion-weighted PROPELLER technique to quantitative assessment of the spinal cord immediately adjacent to metallic spinal fusion instrumentation. METHODS: Morphological and diffusion-weighted MRI of the spinal cord was collected from 5 subjects with implanted cervical spinal fusion hardware. Conventional and multispectral diffusion-weighted images were also collected on a normative non-instrumented control cohort and utilized for methodological stability analysis. Variance of the ADC values derived from the normative control group was then analyzed on a subject-by-subject basis and qualitatively correlated with clinical morphological interpretations. RESULTS: Normative control ADC values within the spinal cord were stable across DWI methods for a b value of 600 s/mm2, though this stability degraded at lower b value levels. Susceptibility artifacts precluded conventional DWI analysis of the cord in subjects with spinal fusion hardware in 4 of the 5 test cases. On the contrary, multispectral PROPELLER DWI produced viable ADC measurements within the cord of all 5 instrumented subjects. Instrumented cord regions without obvious pathology (N = 4) showed ADC values that were lower than expected, whereas one subject with diagnosed myelomalacia showed abnormally elevated ADC. CONCLUSIONS: In the absence of instrumentation, multispectral DWI provides quantitative capabilities that match with those of conventional DWI approaches. In a preliminary instrumented subject analysis, cord ADC values showed both expected and unexpected variations from the normative cohort. These slides can be retrieved under Electronic Supplementary Material.

3D-multi-spectral T2 mapping near metal implants.

PURPOSE: Due to host-mediated adverse reaction to metallic debris, there is an increasing need for noninvasive assessment of the soft tissue surrounding large joint arthroplasties. Quantitative T2 mapping can be beneficial for tissue characterization and early diagnosis of tissue pathology but current T2 mapping techniques lack the capability to image near metal hardware. A novel multi-spectral T2 mapping technique is proposed to address this unmet need. METHODS: A T2 mapping pulse sequence based on routinely implemented 3D multi-spectral imaging (3D-MSI) pulse sequences is described and demonstrated. The 3D-MSI pulse sequence is altered to acquire images at 2 echo times. Phantom and knee experiments were performed to assess the quantitative capabilities of the sequence in comparison to a commercially available T2 mapping sequence. The technique was demonstrated for use within a clinical protocol in 2 total hip arthroplasty (THA) cases to assess T2 variations within the periprosthetic joint space. RESULTS: The proposed multi-spectral T2 mapping technique agreed, within experimental errors, with T2 values derived from a commercially available clinical standard of care T2 mapping sequence. The same level of agreement was observed in quantitative phantoms and in vivo experiments. In THA cases, the method was able to assess variations of T2 within the synovial envelope immediately adjacent to implant interfaces. CONCLUSIONS: The proposed 3D-MSI T2 mapping sequence was successfully demonstrated in assessing tissue T2 variations near metal implants.

Multispectral diffusion-weighted imaging near metal implants.

PURPOSE: The need for diffusion-weighted-imaging (DWI) near metallic implants is becoming increasingly relevant for a variety of clinical diagnostic applications. Conventional DWI methods are significantly hindered by metal-induced image artifacts. A novel approach relying on multispectral susceptibility artifact reduction techniques is presented to address this unmet need. METHODS: DWI near metal implants is achieved through a combination of several advanced MRI acquisition technologies. Previously described approaches to Carr-Purcell-Meiboom-Gill spin-echo train DWI sequences using the periodically rotated overlapping parallel lines with enhanced reconstruction are combined with multispectral-imaging metal artifact reduction principles to provide DWI with substantially reduced artifact levels. The presented methods are applied to limited sets of slices over areas of sarcoma risk near six implanted devices. RESULTS: Using the presented methods, DWI assessment without bulk image distortions is demonstrated in the immediate vicinity of metallic interfaces. In one subject, the apparent diffusion coefficient was reduced in a region of suspected sarcoma directly adjacent to fixation hardware. CONCLUSIONS: An initial demonstration of minimal-artifact multispectral DWI in the near vicinity of metallic hardware is described and successfully demonstrated on clinical subjects. Magn Reson Med 79:987-993, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

Accelerated whole-brain multi-parameter mapping using blind compressed sensing.

PURPOSE: To introduce a blind compressed sensing (BCS) framework to accelerate multi-parameter MR mapping, and demonstrate its feasibility in high-resolution, whole-brain T1ρ and T2 mapping. METHODS: BCS models the evolution of magnetization at every pixel as a sparse linear combination of bases in a dictionary. Unlike compressed sensing, the dictionary and the sparse coefficients are jointly estimated from undersampled data. Large number of non-orthogonal bases in BCS accounts for more complex signals than low rank representations. The low degree of freedom of BCS, attributed to sparse coefficients, translates to fewer artifacts at high acceleration factors (R). RESULTS: From 2D retrospective undersampling experiments, the mean square errors in T1ρ and T2 maps were observed to be within 0.1% up to R = 10. BCS was observed to be more robust to patient-specific motion as compared to other compressed sensing schemes and resulted in minimal degradation of parameter maps in the presence of motion. Our results suggested that BCS can provide an acceleration factor of 8 in prospective 3D imaging with reasonable reconstructions. CONCLUSION: BCS considerably reduces scan time for multiparameter mapping of the whole brain with minimal artifacts, and is more robust to motion-induced signal changes compared to current compressed sensing and principal component analysis-based techniques.

Sparse spectral deconvolution algorithm for noncartesian MR spectroscopic imaging.

PURPOSE: To minimize line shape distortions and spectral leakage artifacts in MR spectroscopic imaging (MRSI). METHODS: A spatially and spectrally regularized non-Cartesian MRSI algorithm that uses the line shape distortion priors, estimated from water reference data, to deconvolve the spectra is introduced. Sparse spectral regularization is used to minimize noise amplification associated with deconvolution. A spiral MRSI sequence that heavily oversamples the central k-space regions is used to acquire the MRSI data. The spatial regularization term uses the spatial supports of brain and extracranial fat regions to recover the metabolite spectra and nuisance signals at two different resolutions. Specifically, the nuisance signals are recovered at the maximum resolution to minimize spectral leakage, while the point spread functions of metabolites are controlled to obtain acceptable signal-to-noise ratio. RESULTS: The comparisons of the algorithm against Tikhonov regularized reconstructions demonstrates considerably reduced line-shape distortions and improved metabolite maps. CONCLUSION: The proposed sparsity constrained spectral deconvolution scheme is effective in minimizing the line-shape distortions. The dual resolution reconstruction scheme is capable of minimizing spectral leakage artifacts.

Intravoxel incoherent motion (IVIM) detects femoral head ischemia in a piglet model of Legg-Calvé-Perthes disease.

There is a clinical need for alternatives to gadolinium contrast-enhanced magnetic resonance imaging (MRI) to facilitate early detection and assessment of femoral head ischemia in pediatric patients with Legg-Calvé-Perthes disease (LCPD), a juvenile form of idiopathic osteonecrosis of the femoral head. The purpose of this study was to determine if intravoxel incoherent motion (IVIM), a noncontrast-enhanced MRI method to simultaneously measure tissue perfusion and diffusion, can detect femoral head ischemia using a piglet model of LCPD. Twelve 6-week-old piglets underwent unilateral hip surgery to induce complete femoral head ischemia. The unoperated, contralateral femoral head served as a perfused control. The bilateral hips of the piglets were imaged in vivo at 3T MRI using IVIM and contrast-enhanced MRI 1 week after surgery. Median apparent diffusion coefficient (ADC) and IVIM parameters (diffusion coefficient: Ds; perfusion coefficient: Df; perfusion fraction: f; and perfusion flux: f*Df) were compared between regions of interest comprising the epiphyseal bone marrow of the ischemic and control femoral heads. Contrast-enhanced MRI confirmed complete femoral head ischemia in 11/12 piglets. IVIM perfusion fraction (f) and flux (f*Df) were significantly decreased in the ischemic versus control femoral heads: on average, f decreased 47 ± 27% (Δf = -0.055 ± 0.034; p = 0.0003) and f*Df decreased 50 ± 27% (Δf*Df = -0.59 ± 0.49 × 10-3 mm2/s; p = 0.0026). In contrast, IVIM diffusion coefficient (Ds) and ADC were significantly increased in the ischemic versus control femoral heads: on average, Ds increased 78 ± 21% (ΔDs = 0.60 ± 0.14 × 10-3 mm2/s; p < 0.0001) and ADC increased 60 ± 36% (ΔADC = 0.50 ± 0.23 × 10-3 mm2/s; p < 0.0001). In conclusion, IVIM is sensitive in detecting bone marrow ischemia in a piglet model of LCPD.

Architectural Distortion on Screening Digital Breast Tomosynthesis: Pathologic Outcomes and Indicators of Malignancy.

OBJECTIVE: Digital breast tomosynthesis (DBT) has significantly improved cancer detection capabilities through its identification of subtle findings often imperceptible on 2D digital mammography, particularly architectural distortion (AD). The purpose of this study was to analyze of suspicious AD detected on screening DBT to evaluate the incidence of malignancy and to determine other patient or imaging characteristics in these cases as possible predictors of malignancy. METHODS: This was an IRB approved retrospective analysis of subjects with AD detected on DBT screening mammography who were given a biopsy recommendation between January 1, 2016, and June 30, 2018. Univariate analysis of various imaging characteristics and patient high-risk factors was performed for statistical correlation with diagnosis of malignancy. RESULTS: In the 218 DBT-detected AD findings with a final BI-RADS assessment of 4 or 5 on diagnostic workup, 94 (43.1%) yielded malignancy, 57 (26.2%) were classified as high-risk, and 67 (30.7%) were benign. There was a strong statistically significant association with malignancy in the cases with an US correlate (P < 0.0001). There was a statistically significant inverse correlation between malignancy and one-view findings (P = 0.0002). The presence of AD on 2D (P = 0.005) or synthetic 2D views (P = 0.002) showed statistically significant correlations with malignancy, whereas breast density or high-risk factors (P = 0.316) did not. CONCLUSION: AD detected on DBT that persists on further workup and has no explainable cause should be considered suspicious for malignancy. Identification of the AD on both standard mammographic views and the presence of an US correlate significantly increase the probability of malignancy.

Analysis and Evaluation of a Deep Learning Reconstruction Approach with Denoising for Orthopedic MRI.

PURPOSE: To evaluate two settings (noise reduction of 50% or 75%) of a deep learning (DL) reconstruction model relative to each other and to conventional MR image reconstructions on clinical orthopedic MRI datasets. MATERIALS AND METHODS: This retrospective study included 54 patients who underwent two-dimensional fast spin-echo MRI for hip (n = 22; mean age, 44 years ± 13 [standard deviation]; nine men) or shoulder (n = 32; mean age, 56 years ± 17; 17 men) conditions between March 2019 and June 2020. MR images were reconstructed with conventional methods and the vendor-provided and commercially available DL model applied with 50% and 75% noise reduction settings (DL 50 and DL 75, respectively). Quantitative analytics, including relative anatomic edge sharpness, relative signal-to-noise ratio (rSNR), and relative contrast-to-noise ratio (rCNR) were computed for each dataset. In addition, the image sets were randomized, blinded, and presented to three board-certified musculoskeletal radiologists for ranking based on overall image quality and diagnostic confidence. Statistical analysis was performed with a nonparametric hypothesis comparing derived quantitative metrics from each reconstruction approach. In addition, inter- and intrarater agreement analysis was performed on the radiologists' rankings. RESULTS: Both denoising settings of the DL reconstruction showed improved edge sharpness, rSNR, and rCNR relative to the conventional reconstructions. The reader rankings demonstrated strong agreement, with both DL reconstructions outperforming the conventional approach (Gwet agreement coefficient = 0.98). However, there was lower agreement between the readers on which DL reconstruction denoising setting produced higher-quality images (Gwet agreement coefficient = 0.31 for DL 50 and 0.35 for DL 75). CONCLUSION: The vendor-provided DL MRI reconstruction showed higher edge sharpness, rSNR, and rCNR in comparison with conventional methods; however, optimal levels of denoising may need to be further assessed.Keywords: MRI Reconstruction Method, Deep Learning, Image Analysis, Signal-to-Noise Ratio, MR-Imaging, Neural Networks, Hip, Shoulder, Physics, Observer Performance, Technology Assessment Supplemental material is available for this article. © RSNA, 2021.

Optimization of hyperparameters for SMS reconstruction.

PURPOSE: Simultaneous multi-slice (SMS) imaging accelerates MRI data acquisition by exciting multiple image slices with a single radiofrequency pulse. Overlapping slices encoded in acquired signal are separated using a mathematical model, which requires estimation of image reconstruction kernels using calibration data. Several parameters used in SMS reconstruction impact the quality and fidelity of final images. Therefore, finding an optimal set of reconstruction parameters is critical to ensure that accelerated acquisition does not significantly degrade resulting image quality. METHODS: Gradient-echo echo planar imaging data were acquired with a range of SMS acceleration factors from a cohort of five volunteers with no known neurological pathology. Images were collected using two available phased-array head coils (a 48-channel array and a reduced diameter 32-channel array) that support SMS. Data from these coils were identically reconstructed offline using a range of coil compression factors and reconstruction kernel parameters. A hybrid space (k-x), externally-calibrated coil-by-coil slice unaliasing approach was used for image reconstruction. The image quality of the resulting reconstructed SMS images was assessed by evaluating correlations with identical echo-planar reference data acquired without SMS. A finger tapping functional MRI (fMRI) experiment was also performed and group analysis results were compared between data sets reconstructed with different coil compression levels. RESULTS: Between the two RF coils tested in this study, the 32-channel coil with smaller dimensions clearly outperformed the larger 48-channel coil in our experiments. Generally, a large calibration region (144-192 samples) and small kernel sizes (2-4 samples) in ky direction improved image quality. Use of regularization in the kernel fitting procedure had a notable impact on the fidelity of reconstructed images and a regularization value 0.0001 provided good image quality. With optimal selection of other hyperparameters in the hybrid space SMS unaliasing algorithm, coil compression caused small reduction in correlation between single-band and SMS unaliased images. Similarly, group analysis of fMRI results did not show a significant influence of coil compression on resulting image quality. CONCLUSIONS: This study demonstrated that the hyperparameters used in SMS reconstruction need to be fine-tuned once the experimental factors such as the RF receive coil and SMS factor have been determined. A cursory evaluation of SMS reconstruction hyperparameter values is therefore recommended before conducting a full-scale quantitative study using SMS technologies.

Blind Compressed Sensing Enables 3-Dimensional Dynamic Free Breathing Magnetic Resonance Imaging of Lung Volumes and Diaphragm Motion.

OBJECTIVES: The objective of this study was to increase the spatial and temporal resolution of dynamic 3-dimensional (3D) magnetic resonance imaging (MRI) of lung volumes and diaphragm motion. To achieve this goal, we evaluate the utility of the proposed blind compressed sensing (BCS) algorithm to recover data from highly undersampled measurements. MATERIALS AND METHODS: We evaluated the performance of the BCS scheme to recover dynamic data sets from retrospectively and prospectively undersampled measurements. We also compared its performance against that of view-sharing, the nuclear norm minimization scheme, and the l1 Fourier sparsity regularization scheme. Quantitative experiments were performed on a healthy subject using a fully sampled 2D data set with uniform radial sampling, which was retrospectively undersampled with 16 radial spokes per frame to correspond to an undersampling factor of 8. The images obtained from the 4 reconstruction schemes were compared with the fully sampled data using mean square error and normalized high-frequency error metrics. The schemes were also compared using prospective 3D data acquired on a Siemens 3 T TIM TRIO MRI scanner on 8 healthy subjects during free breathing. Two expert cardiothoracic radiologists (R1 and R2) qualitatively evaluated the reconstructed 3D data sets using a 5-point scale (0-4) on the basis of spatial resolution, temporal resolution, and presence of aliasing artifacts. RESULTS: The BCS scheme gives better reconstructions (mean square error = 0.0232 and normalized high frequency = 0.133) than the other schemes in the 2D retrospective undersampling experiments, producing minimally distorted reconstructions up to an acceleration factor of 8 (16 radial spokes per frame). The prospective 3D experiments show that the BCS scheme provides visually improved reconstructions than the other schemes do. The BCS scheme provides improved qualitative scores over nuclear norm and l1 Fourier sparsity regularization schemes in the temporal blurring and spatial blurring categories. The qualitative scores for aliasing artifacts in the images reconstructed by nuclear norm scheme and BCS scheme are comparable.The comparisons of the tidal volume changes also show that the BCS scheme has less temporal blurring as compared with the nuclear norm minimization scheme and the l1 Fourier sparsity regularization scheme. The minute ventilation estimated by BCS for tidal breathing in supine position (4 L/min) and the measured supine inspiratory capacity (1.5 L) is in good correlation with the literature. The improved performance of BCS can be explained by its ability to efficiently adapt to the data, thus providing a richer representation of the signal. CONCLUSION: The feasibility of the BCS scheme was demonstrated for dynamic 3D free breathing MRI of lung volumes and diaphragm motion. A temporal resolution of ∼500 milliseconds, spatial resolution of 2.7 × 2.7 × 10 mm, with whole lung coverage (16 slices) was achieved using the BCS scheme.

A variable splitting based algorithm for fast multi-coil blind compressed sensing MRI reconstruction.

Recent work on blind compressed sensing (BCS) has shown that exploiting sparsity in dictionaries that are learnt directly from the data at hand can outperform compressed sensing (CS) that uses fixed dictionaries. A challenge with BCS however is the large computational complexity during its optimization, which limits its practical use in several MRI applications. In this paper, we propose a novel optimization algorithm that utilize variable splitting strategies to significantly improve the convergence speed of the BCS optimization. The splitting allows us to efficiently decouple the sparse coefficient, and dictionary update steps from the data fidelity term, resulting in subproblems that take closed form analytical solutions, which otherwise require slower iterative conjugate gradient algorithms. Through experiments on multi coil parametric MRI data, we demonstrate the superior performance of BCS over conventional CS schemes, while achieving convergence speed up factors of over 10 fold over the previously proposed implementation of the BCS algorithm.