MRI Reconstruction Research Articles

Federated Learning of Generative Image Priors for MRI Reconstruction.

Multi-institutional efforts can facilitate training of deep MRI reconstruction models, albeit privacy risks arise during cross-site sharing of imaging data. Federated learning (FL) has recently been introduced to address privacy concerns by enabling distributed training without transfer of imaging data. Existing FL methods employ conditional reconstruction models to map from undersampled to fully-sampled acquisitions via explicit knowledge of the accelerated imaging operator. Since conditional models generalize poorly across different acceleration rates or sampling densities, imaging operators must be fixed between training and testing, and they are typically matched across sites. To improve patient privacy, performance and flexibility in multi-site collaborations, here we introduce Federated learning of Generative IMage Priors (FedGIMP) for MRI reconstruction. FedGIMP leverages a two-stage approach: cross-site learning of a generative MRI prior, and prior adaptation following injection of the imaging operator. The global MRI prior is learned via an unconditional adversarial model that synthesizes high-quality MR images based on latent variables. A novel mapper subnetwork produces site-specific latents to maintain specificity in the prior. During inference, the prior is first combined with subject-specific imaging operators to enable reconstruction, and it is then adapted to individual cross-sections by minimizing a data-consistency loss. Comprehensive experiments on multi-institutional datasets clearly demonstrate enhanced performance of FedGIMP against both centralized and FL methods based on conditional models.

Interosseous tendon inflammation in the hands of patients with clinically suspect arthralgia: analysis of MRI data from a prospective cohort study.

Inflammation around the tendons of interosseous muscles of the hand (interosseous tendon inflammation) was recently observed with MRI for the first time in patients with rheumatoid arthritis and in at-risk individuals with detectable anti-citrullinated protein antibodies, generating the hypothesis that interosseous tendon inflammation precedes clinical arthritis. To better understand the role of interosseous tendon inflammation during the development of rheumatoid arthritis, we studied the frequency of interosseous tendon inflammation in healthy individuals and in those with arthralgia that was suspected of progressing to rheumatoid arthritis (ie, clinically suspect arthralgia) and the association of interosseous tendon inflammation with other symptoms of inflamed joint tissues and with clinical arthritis development. Adult (age ≥18 years) patients who presented with clinically suspect arthralgia and symptom-free (control) individuals underwent contrast-enhanced hand MRI. MRIs were evaluated for interosseous tendon inflammation on the radial and ulnar sides of the second to fifth metacarpophalangeal joints, and for synovitis, tenosynovitis, and osteitis using the rheumatoid arthritis MRI scoring system. Patients with clinically suspect arthralgia were followed up for clinical arthritis development. The presence of local tenosynovium was examined using immunohistochemistry for anti-CD55 and anti-CD68 on tissue from the hands of three embalmed bodies donated for scientific research. The primary outcome for the cross-sectional part of the study was the presence of interosseous tendon inflammation on MRI. The primary outcome for the longitudinal part of the study was development of clinical arthritis. Between April 3, 2012, and May 20, 2020, 667 patients with clinically suspect arthralgia (mean age 44 years [SD 13], 504 [76%] were women and 163 [24%] were men) underwent contrast-enhanced hand MRI. Between Nov 1, 2013, and Nov 30, 2014, 193 symptom-free controls were recruited (mean age 50 years [SD 16], 136 [70%] were women and 57 [30%] were men). Two (1%) of 193 symptom-free controls had interosseous tendon inflammation. Immunohistochemistry of cadaveric hand tissues showed no tenosynovium surrounding interosseous tendons. At inclusion, 67 (10%) of 667 patients with clinically suspect arthralgia had interosseous tendon inflammation (p<0·0001 vs symptom-free controls). Interosseous tendon inflammation occurred more frequently if synovitis (odds ratio [OR] 2·2 [95% CI 1·2-4·2]), or tenosynovitis (OR 9·7 [5·5-17·0]), was present at metacarpophalangeal joints. A three-dimensional MRI reconstruction suggested confluency of interosseous tendon inflammation with metacarpophalangeal-flexor-tenosynovitis. 91 (16%) of 558 patients with clinically suspect arthralgia developed clinical arthritis during follow-up (median total follow-up 25·3 months [95% CI 25·1-25·5]). Patients with clinically suspect arthralgia with interosseous tendon inflammation had a higher risk of developing clinical arthritis (hazard ratio [HR] 4·5 [2·8-7·2]), which was attenuated but still significant after adjusting for concomitant synovitis, tenosynovitis, or osteitis (HR 1·7 [1·02-2·8]). Interosseous tendon inflammation is almost absent in symptom-free individuals but occurs in people with clinically suspect arthralgia, in whom it correlates with symptoms and is associated with the development of clinical arthritis. The absence of local tenosynovium suggests that interosseous tendon inflammation arises from expanding local subclinical inflammation in the pre-arthritis phase of rheumatoid arthritis. European Research Council and the Dutch Arthritis Society.

Evaluating efficient SENSE algorithms to deblur spiral MRI with fat/water separation.

The combination of SENSE and spiral imaging with fat/water separation enables high temporal efficiency. However, the corresponding computation increases due to the blurring/deblurring operation across the multi-channel data. This study presents two alternative models to simplify computational complexity in the original full model (model 1). The performances of the models are evaluated in terms of the computation time and reconstruction error. Two approximated spiral MRI reconstruction models were proposed: the comprehensive blurring before coil operation (model 2) and the regional blurring before coil operation (model 3), respectively, by altering the order of coil-sensitivity encoding process to distribute signals among the multi-channel coils. Four subjects were recruited for scanning both fully sampled T1 - and T2 -weighted brain image data with simulated undersampling for testing the computational efficiency and accuracy on the approximation models. Based on the examples, the computation time can be reduced to 31%-47% using model 2, and to 39%-56% using model 3.The quality of the water image remains unchanged among the three models, whereas the primary difference in image quality is in the fat channel. The fat images from model 3 are consistent with those from model 1, but those from model 2 have higher normalized error, differing by up to 4.8%. Model 2 provides the fastest computation but exhibits higher error in the fat channel, particularly in the high field and with long acquisition window. Model 3, an abridged alternative, is also faster than the full model and can maintain high accuracy in reconstruction.

Data-efficient Bayesian learning for radial dynamic MR reconstruction.

Cardiac MRI has become the gold-standard imaging technique for assessing cardiovascular morphology and function. In spite of this, its slow data acquisition process presents imaging challenges due to the motion from heartbeats, respiration, and blood flow. In recent studies, deep learning (DL) algorithms have shown promising results for the task of image reconstruction. However, there have been instances where they have introduced artifacts that may be misinterpreted as pathologies or may obscure the detection of pathologies. Therefore, it is important to obtain a metric, such as the uncertainty of the network output, that identifies such artifacts. However, this can be quite challenging for large-scale image reconstruction problems such as dynamic multi-coil non-CartesianMRI. To efficiently quantify uncertainties of a physics-informed DL-based image reconstruction method for a large-scale accelerated 2D multi-coil dynamic radial MRI reconstruction problem, and demonstrate the benefits of physics-informed DL over model-agnostic DL in reducing uncertainties while at the same time improving imagequality. We extended a recently proposed physics-informed 2D U-Net that learns spatio-temporal slices (named XT-YT U-Net), and employed it for the task of uncertainty quantification (UQ) by using Monte Carlo dropout and a Gaussian negative log-likelihood loss function. Our data comprised 2D dynamic MR images acquired with a radial balanced steady-state free precession sequence. The XT-YT U-Net, which allows for training with a limited amount of data, was trained and validated on a dataset of 15 healthy volunteers, and further tested on data from four patients. An extensive comparison between physics-informed and model-agnostic neural networks (NNs) concerning the obtained image quality and uncertainty estimates was performed. Further, we employed calibration plots to assess the quality of theUQ. The inclusion of the MR-physics model of data acquisition as a building block in the NN architecture led to higher image quality (NRMSE: , PSNR: , and SSIM: ), lower uncertainties ( ), and, based on the calibration plots, an improved UQ compared to its model-agnostic counterpart. Furthermore, the UQ information can be used to differentiate between anatomical structures (e.g., coronary arteries, ventricle boundaries) andartifacts. Using an XT-YT U-Net, we were able to quantify uncertainties of a physics-informed NN for a high-dimensional and computationally demanding 2D multi-coil dynamic MR imaging problem. In addition to improving the image quality, embedding the acquisition model in the network architecture decreased the reconstruction uncertainties as well as quantitatively improved the UQ. The UQ provides additional information to assess the performance of different networkapproaches.

Utility of accelerated T2-weighted turbo spin-echo imaging with deep learning reconstruction in female pelvic MRI: a multi-reader study.

To determine the clinical feasibility of T2-weighted turbo spin-echo (T2-TSE) imaging with deep learning reconstruction (DLR) in female pelvic MRI compared with conventional T2 TSE in terms of image quality and scan time. Between May 2021 and September 2021, 52 women (mean age, 44years ± 12) who underwent 3-T pelvic MRI with additional T2-TSE using a DLR algorithm were included in this single-center prospective study with patient's informed consents. Conventional, DLR, and DLR T2-TSE images with reduced scan times were independently assessed and compared by four radiologists. The overall image quality, differentiation of anatomic details, lesion conspicuity, and artifacts were evaluated using a 5-point scale. Inter-observer agreement of the qualitative scores was compared and reader protocol preferences were then evaluated. In the qualitative analysis of all readers, fast DLR T2-TSE showed significantly better overall image quality, differentiation of anatomic regions, lesion conspicuity, and lesser artifacts than conventional T2-TSE and DLR T2-TSE, despite approximately 50% reduction in scan time (all p < 0.05). The inter-reader agreement for the qualitative analysis was moderate to good. All readers preferred DLR over conventional T2-TSE regardless of scan time and preferred fast DLR T2-TSE (57.7-78.8%), except for one who preferred DLR over fast DLR T2-TSE (53.8% vs. 46.1%). In female pelvic MRI, image quality and accelerated image acquisition for T2-TSE can be significantly improved by using DLR compared to conventional T2-TSE. Fast DLR T2-TSE was non-inferior to DLR T2-TSE in terms of reader preference and image quality. DLR of T2-TSE in female pelvic MRI enables fast imaging along with maintaining optimal image quality compared with parallel imaging-based conventional T2-TSE. • Conventional T2 turbo spin-echo based on parallel imaging has limitations for accelerated image acquisition while maintaining good image quality. • Deep learning image reconstruction showed better image quality in both images obtained using the same or accelerated image acquisition parameters compared with conventional T2 turbo spin-echo in female pelvic MRI. • Deep learning image reconstruction enables accelerated image acquisition while maintaining good image quality in the T2-TSE of female pelvic MRI.

Comparison of the femoral condylar ellipse line and the surgical epicondylar axis: 3D measurement by MRI scans in healthy Chinese people.

The sagittal shapes of the femoral condyles were thought to consist of circles. However, the line connecting the centers of circles was not consistent with the surgical epicondylar axis (SEA) which was commonly used in surgery. Recently, ellipses have been proposed as an alternative method to represent the sagittal femoral condylar shape. Does the condylar ellipse line (CEL) coincide with the SEA in 3D MRI reconstruction analysis? From May to August 2021, a total of 80 healthy subjects were scanned by MRI on the right knee in this retrospective study. The ellipses on the most distal slices of the medial and lateral condyles were determined. A line connecting the centers of the medial and lateral ellipses was the CEL. A line connecting the deepest point of the medial sulcus and the most prominent point of the lateral epicondyle was the SEA. Angular measurement of the SEA and the CEL relative to the posterior condylar line (PCL) and the distal condylar line (DCL) was performed on an axial and coronal view of the 3D model, respectively. Measurements were compared between males and females by using the independent-samples t-test. Pearson correlation was used to analyze the relationship between SEA-PCL and CEL-PCL, SEA-DCL, and CEL-DCL. On the axial view, the mean SEA-CEL was 0.35° ± 0.96°. SEA-PCL (2.91° ± 1.40°) had a high correlation with CEL-PCL (3.27° ± 1.11°) (r = 0.731, p < 0.001). On the coronal view, the mean coronal SEA-CEL was 1.35° ± 1.13°. SEA-DCL (1.35° ± 1.13°) had a low correlation with CEL-DCL (0.18° ± 0.84°) (r = 0.319, p = 0.007). On the sagittal view, the outlet points of the CEL on the medial and lateral epicondyles were anatomically located in the anteroinferior direction to the SEA. CEL traversed the medial and lateral epicondyles, which has a mean deviation of 0.35° with SEA on axial view and a mean deviation of 0.18° with DCL on coronal view. This study suggested that the ellipse approach is an improved scheme for representing the femoral condylar shape.

Clinical Impact of Deep Learning Reconstruction in MRI

Deep learning has been recognized as a paradigm-shifting tool in radiology. Deep learning reconstruction (DLR) has recently emerged as a technology used in the image reconstruction process of MRI, which is an essential procedure in generating MR images. Denoising, which is the first DLR application to be realized in commercial MRI scanners, improves signal-to-noise ratio. When applied to lower magnetic field-strength scanners, the signal-to-noise ratio can be increased without extending the imaging time, and image quality is comparable to that of higher-field-strength scanners. Shorter imaging times decrease patient discomfort and reduce MRI scanner running costs. The incorporation of DLR into accelerated acquisition imaging techniques, such as parallel imaging or compressed sensing, shortens the reconstruction time. DLR is based on supervised learning using convolutional layers and is divided into the following three categories: image domain, k-space learning, and direct mapping types. Various studies have reported other derivatives of DLR, and several have shown the feasibility of DLR in clinical practice. Although DLR efficiently reduces Gaussian noise from MR images, denoising makes image artifacts more prominent, and a solution to this problem is desired. Depending on the training of the convolutional neural network, DLR may change the imaging features of lesions and obscure small lesions. Therefore, radiologists may need to adopt the habit of questioning whether any information has been lost on images that appear clean. ©RSNA, 2023 Quiz questions for this article are available in the supplemental material.

Dual-domain accelerated MRI reconstruction using transformers with learning-based undersampling.

Acceleration in MRI has garnered much attention from the deep-learning community in recent years, particularly for imaging large anatomical volumes such as the abdomen or moving targets such as the heart. A variety of deep learning approaches have been investigated, with most existing works using convolutional neural network (CNN)-based architectures as the reconstruction backbone, paired with fixed, rather than learned, k-space undersampling patterns. In both image domain and k-space, CNN-based architectures may not be optimal for reconstruction due to its limited ability to capture long-range dependencies. Furthermore, fixed undersampling patterns, despite ease of implementation, may not lead to optimal reconstruction. Lastly, few deep learning models to date have leveraged temporal correlation across dynamic MRI data to improve reconstruction. To address these gaps, we present a dual-domain (image and k-space), transformer-based reconstruction network, paired with learning-based undersampling that accepts temporally correlated sequences of MRI images for dynamic reconstruction. We call our model DuDReTLU-net. We train the network end-to-end against fully sampled ground truth dataset. Human cardiac CINE images undersampled at different factors (5-100) were tested. Reconstructed images were assessed both visually and quantitatively via the structural similarity index, mean squared error, and peak signal-to-noise. Experimental results show superior performance of DuDReTLU-net over state-of-the-art methods (LOUPE, k-t SLR, BM3D-MRI) in accelerated MRI reconstruction; ablation studies show that transformer-based reconstruction outperformed CNN-based reconstruction in both image domain and k-space; dual-domain reconstruction architectures outperformed single-domain reconstruction architectures regardless of reconstruction backbone (CNN or transformer); and dynamic sequence input leads to more accurate reconstructions than single frame input. We expect our results to encourage further research in the use of dual-domain architectures, transformer-based architectures, and learning-based undersampling, in the setting of accelerated MRI reconstruction. The code for this project is made freely available at https://github.com/william2343/dual-domain-mri-recon-nets (Hong et al., 2022).

Brain reserve is associated with cognition in AD and MCI patients with different vascualr risk factors status

AbstractBackgroundDementia due to Alzheimer’s disease (AD) is the major type of dementias in the world. In addition to amyloid plaques and neurofibrillary tangles, cerebrovascular disease is another important pathophysiological cause of AD. Cerebrovascular disease is highly associated with vascular risk factors (VRFs). Well‐treated VRFs might decrease the risk and progression of AD. However, findings on the impact of VRFs treatment status on cognitive decline after the diagnosis of AD established are unclear, and there is no direct evidence supporting well‐treated VRFs can slow down the AD progression.MethodThis study recruited mild cognitive impairment (MCI) and mild AD patients (CDR = 0.5 or 1) with at least one VRF from at Taichung Veteran General Hospital. The treatment status of the VRFs was identified and all participants were evaluated by a standard neuropsychological test battery. We collected the participants’ brain MRI done within one year and performed Apolipoprotein E genotyping. FreeSurfer was used for structural MRI reconstruction and calculated total gray matter volume, hippocampus volume, and AD signature cortical thickness. Cerebrovascular lesions, including perivascular spaces in centrum semiovale, basal ganglia and presence of lobar cerebral microbleeds (LCMBs), and the volume of white matter hyperintensities were estimated by visual rating scales and Lesion Segmentation Tool for Statistical Parametric Mapping. We analyzed the relationships among the treatment status of VRFs, neuropsychological tests, and brain MRI.ResultWe recruited 61 patients in this study. Four patients refused to keep involving the study, 6 patients were not compatible with the criteria of MCI and AD after neuropsychiatric testing, and 4 patients’ brain MRI images cannot be analyzed after quality checked. The remaining 47 patients (15 MCI patients and 32 AD patients) constituted the final sample for analysis. The patients with well‐controlled VRFs present with similar cognitive function compared to the patients with partially‐controlled VRFs even though the more severe hippocampal atrophy was noted in the patients with well controlled VRFs. The severity of white matter hyperintensities and other cerebrovascular lesions are not correlated with the cognitive function in MCI and AD patients.ConclusionBrain reserve in the AD and MCI patients with well‐controlled VRFs might contribute cognitive function

Mesiotemporal Volumetry, Cortical Thickness, and Neuropsychological Deficits in the Long-term Course of Limbic Encephalitis.

Limbic encephalitis (LE) is an autoimmune disease often associated with temporal lobe epilepsy and subacute memory deficits. It is categorized into serologic subgroups, which differ in clinical progress, therapy response, and prognosis. Using longitudinal MRI analysis, we hypothesized that mesiotemporal and cortical atrophy rates would reveal serotype-specific patterns and reflect disease severity. In this longitudinal case-control study, all individuals with antibody-positive (glutamic acid decarboxylase 65 [GAD], leucine-rich glioma-inactivated protein 1 [LGI1], contactin-associated protein 2 [CASPR2], and N-methyl-d-aspartate receptor [NMDAR]) nonparaneoplastic LE according to Graus' diagnostic criteria treated between 2005 and 2019 at the University Hospital Bonn were enrolled. A longitudinal healthy cohort was included as the control group. Subcortical segmentation and cortical reconstruction of T1-weighted MRI were performed using the longitudinal framework in FreeSurfer. We applied linear mixed models to examine mesiotemporal volumes and cortical thickness longitudinally. Two hundred fifty-seven MRI scans from 59 individuals with LE (34 female, age at disease onset [mean ± SD] 42.5 ± 20.4 years; GAD: n = 30, 135 scans; LGI1: n = 15, 55 scans; CASPR2: n = 9, 37 scans; and NMDAR: n = 5, 30 scans) were included. The healthy control group consisted of 128 scans from 41 individuals (22 female, age at first scan [mean ± SD] 37.7 ± 14.6 years). The amygdalar volume at disease onset was significantly higher in individuals with LE (p ≤ 0.048 for all antibody subgroups) compared with that in healthy controls and decreased over time in all antibody subgroups, except in the GAD subgroup. We observed a significantly higher hippocampal atrophy rate in all antibody subgroups compared with that in healthy controls (all p ≤ 0.002), except in the GAD subgroup. Cortical atrophy rates exceeded normal aging in individuals with impaired verbal memory, while those who were not impaired did not differ significantly from healthy controls. Our data depict higher mesiotemporal volumes in the early disease stage, most likely due to edematous swelling, followed by volume regression and atrophy/hippocampal sclerosis in the late disease stage. Our study reveals a continuous and pathophysiologically meaningful trajectory of mesiotemporal volumetry across all serogroups and provides evidence that LE should be considered a network disorder in which extratemporal involvement is an important determinant of disease severity.

Study on the evaluation of glenoid bone defects by MRI three-dimensional reconstruction

To investigate the feasibility of MRI three-dimensional (3D) reconstruction model in quantifying glenoid bone defect by comparing with CT 3D reconstruction model measurement. Forty patients with shoulder anterior dislocation who met the selection criteria between December 2021 and December 2022 were admitted as study participants. There were 34 males and 6 females with an average age of 24.8 years (range, 19-32 years). The injury caused by sports injury in 29 cases and collision injury in 6 cases, and 5 cases had no obvious inducement. The time from injury to admission ranged from 4 to 72 months (mean, 28.5 months). CT and MRI were performed on the patients' shoulder joints, and a semi-automatic segmentation of the images was done with 3D slicer software to construct a glenoid model. The length of the glenoid bone defect was measured on the models by 2 physicians. The intra-group correlation coefficient ( ICC) was used to evaluate the consistency between the 2 physicians, and Bland-Altman plots were constructed to evaluate the consistency between the 2 methods. The length of the glenoid bone defects measured on MRI 3D reconstruction model was (3.83±1.36) mm/4.00 (0.58, 6.13) mm for physician 1 and (3.91±1.20) mm/3.86 (1.39, 5.96) mm for physician 2. The length of the glenoid bone defects measured on CT 3D reconstruction model was (3.81±1.38) mm/3.80 (0.60, 6.02) mm for physician 1 and (3.99±1.19) mm/4.00 (1.68, 6.38) mm for physician 2. ICC and Bland-Altman plot analysis showed good consistency. The ICC between the 2 physicians based on MRI and CT 3D reconstruction model measurements were 0.73 [95% CI (0.54, 0.85)] and 0.80 [95% CI (0.65, 0.89)], respectively. The 95% CI of the difference between the two measurements of physicians 1 and 2 were (-0.46, 0.49) and (-0.68, 0.53), respectively. The measurement of glenoid bone defect based on MRI 3D reconstruction model is consistent with that based on CT 3D reconstruction model. MRI can be used instead of CT to measure glenoid bone defects in clinic, and the soft tissue of shoulder joint can be observed comprehensively while reducing radiation.

High-fidelity mesoscale in-vivo diffusion MRI through gSlider-BUDA and circular EPI with S-LORAKS reconstruction

PurposeTo develop a high-fidelity diffusion MRI acquisition and reconstruction framework with reduced echo-train-length for less T2* image blurring compared to typical highly accelerated echo-planar imaging (EPI) acquisitions at sub-millimeter isotropic resolution. MethodsWe first proposed a circular-EPI trajectory with partial Fourier sampling on both the readout and phase-encoding directions to minimize the echo-train-length and echo time. We then utilized this trajectory in an interleaved two-shot EPI acquisition with reversed phase-encoding polarity, to aid in the correction of off-resonance-induced image distortions and provide complementary k-space coverage in the missing partial Fourier regions. Using model-based reconstruction with structured low-rank constraint and smooth phase prior, we corrected the shot-to-shot phase variations across the two shots and recover the missing k-space data. Finally, we combined the proposed acquisition/reconstruction framework with an SNR-efficient RF-encoded simultaneous multi-slab technique, termed gSlider, to achieve high-fidelity 720 µm and 500 µm isotropic resolution in-vivo diffusion MRI. ResultsBoth simulation and in-vivo results demonstrate the effectiveness of the proposed acquisition and reconstruction framework to provide distortion-corrected diffusion imaging at the mesoscale with markedly reduced T2*-blurring. The in-vivo results of 720 µm and 500 µm datasets show high-fidelity diffusion images with reduced image blurring and echo time using the proposed approaches. ConclusionsThe proposed method provides high-quality distortion-corrected diffusion-weighted images with ∼40% reduction in the echo-train-length and T2* blurring at 500µm-isotropic-resolution compared to standard multi-shot EPI.

A CauchyTV non-convex regularization model for MRI reconstruction

A CauchyTV non-convex regularization model for MRI reconstruction

Fast Multi-Contrast MRI Acquisition by Optimal Sampling of Information Complementary to Pre-acquired MRI Contrast.

Recent studies on multi-contrast MRI reconstruction have demonstrated the potential of further accelerating MRI acquisition by exploiting correlation between contrasts. Most of the state-of-the-art approaches have achieved improvement through the development of network architectures for fixed under-sampling patterns, without considering inter-contrast correlation in the under-sampling pattern design. On the other hand, sampling pattern learning methods have shown better reconstruction performance than those with fixed under-sampling patterns. However, most under-sampling pattern learning algorithms are designed for single contrast MRI without exploiting complementary information between contrasts. To this end, we propose a framework to optimize the under-sampling pattern of a target MRI contrast which complements the acquired fully-sampled reference contrast. Specifically, a novel image synthesis network is introduced to extract the redundant information contained in the reference contrast, which is exploited in the subsequent joint pattern optimization and reconstruction network. We have demonstrated superior performance of our learned under-sampling patterns on both public and in-house datasets, compared to the commonly used under-sampling patterns and state-of-the-art methods that jointly optimize the reconstruction network and the under-sampling patterns, up to 8-fold under-sampling factor.

PD-0979 Evaluation of commercial deep learning MRI reconstruction for synthetic CT generation in prostate RT

PD-0979 Evaluation of commercial deep learning MRI reconstruction for synthetic CT generation in prostate RT

Latent signal models: Learning compact representations of signal evolution for improved time‐resolved, multi‐contrast MRI

To improve time-resolved reconstructions by training auto-encoders to learn compact representations of Bloch-simulated signal evolution and inserting the decoder into the forward model. Building on model-based nonlinear and linear subspace techniques, we train auto-encoders on dictionaries of simulated signal evolution to learn compact, nonlinear, latent representations. The proposed latent signal model framework inserts the decoder portion of the auto-encoder into the forward model and directly reconstructs the latent representation. Latent signal models essentially serve as a proxy for fast and feasible differentiation through the Bloch equations used to simulate signal. This work performs experiments in the context of T2 -shuffling, gradient echo EPTI, and MPRAGE-shuffling. We compare how efficiently auto-encoders represent signal evolution in comparison to linear subspaces. Simulation and in vivo experiments then evaluate if reducing degrees of freedom by incorporating our proxy for the Bloch equations, the decoder portion of the auto-encoder, into the forward model improves reconstructions in comparison to subspace constraints. An auto-encoder with 1 real latent variable represents single-tissue fast spin echo, EPTI, and MPRAGE signal evolution to within 0.15% normalized RMS error, enabling reconstruction problems with 3 degrees of freedom per voxel (real latent variable + complex scaling) in comparison to linear models with 4-8 degrees of freedom per voxel. In simulated/in vivo T2 -shuffling and in vivo EPTI experiments, the proposed framework achieves consistent quantitative normalized RMS error improvement over linear approaches. From qualitative evaluation, the proposed approach yields images with reduced blurring and noise amplification in MPRAGE-shuffling experiments. Directly solving for nonlinear latent representations of signal evolution improves time-resolved MRI reconstructions.

Realidad aumentada en cirugía hepato-bilio-pancreática. Una tecnología al alcance de la mano

Augmented reality is a technology that opens new possibilities in surgery. Its implementation in a hepatobiliary-pancreatic surgery unit is presented in relation to preoperative planning, intraoperative support and teaching.For surgical planning, 3D CT and MRI reconstructions have been used to evaluate complex cases, making the interpretation of the anatomy more precise and the planning of the technique simpler.At an intraoperative level, it has allowed remote holographic connection between specialists, the substitution of physical elements for virtual elements, and the use of virtual consultation models and surgical guides.In teaching, new lessons include the retransmission of a surgery with the support of virtual elements for a better understanding by the students.Being the experience satisfactory, augmented reality could be applied in the future of hepatobiliary-pancreatic surgery to improve its results.

Accelerated submillimeter wave-encoded magnetic resonance imaging via deep untrained neural network.

Wave gradient encoding can adequately utilize coil sensitivity profiles to facilitate higher accelerations in parallel magnetic resonance imaging (pMRI). However, there are limitations in mainstream pMRI and a few deep learning (DL) methods for recovering missing data under wave encoding framework: the former is prone to introduce errors from the auto-calibration signals (ACS) signal acquisition and is time-consuming, while the latter requires a large amount of trainingdata. To tackle the above issues, an untrained neural network (UNN) model incorporating wave-encoded physical properties and deep generative model, named WDGM, was proposed with additional ACS- and trainingdata-free. Generally, the proposed method can provide powerful missing data interpolation capability using the wave physical encoding framework and designed UNN to characterize the MR image (k-space data) priors. Specifically, the MRI reconstruction combining physical wave encoding and elaborate UNN is modeled as a generalized minimization problem. The designation of UNN is driven by the coil sensitivity maps (CSM) smoothness and k-space linear predictability. And then, the iterative paradigm to recover the full k-space signal is determined by the projected gradient descent, and the complex computation is unrolled to the network with optimized parameters by the optimizer. Simulated wave encoding and in vivo experiments are exploited to demonstrate the feasibility of the proposed method. The best quantitative metrics RMSE/SSIM/PSNR of 0.0413, 0.9514, and 37.4862 gave competitive results in all experiments with at least six-fold acceleration, respectively. In vivo experiments of human brains and knees showed that the proposed method can achieve comparable reconstruction quality and even has superiority relative to the comparison, especially at a high resolution of 0.67 mm and fewer ACS. In addition, the proposed method has a higher computational efficiency achieving a computation time of 9.6 s/perslice. The model proposed in this work addresses two limitations of MRI reconstruction in the wave encoding framework. The first is to eliminate the need for ACS signal acquisition to perform the time-consuming calibration process and to avoid errors such as motion during the acquisition procedure. Furthermore, the proposed method has clinical application friendly without the need to prepare large training datasets, which is difficult in the clinical. All results of the proposed method demonstrate more confidence in both quantitative and qualitative metrics. In addition, the proposed method can achieve higher computationalefficiency.

Multi-mode dictionaries for fast CS-based dynamic MRI reconstruction

ABSTRACT Dynamic Magnetic Resonance Imaging (dMRI) is a valuable tool for understanding changes in human physiology, but its temporal and spatial resolution can be limited. Compressed sensing (CS) has been used to enhance temporal resolution by acquiring partial k-spaces of each time frame and exploiting sparsity to retain spatial resolution. Invoking CS in dMRI necessitates algorithms that can leverage both spatial sparsity within each time frame and temporal sparsity across time frames. A tensor decomposition-based multi-mode dictionary learning algorithm has been proposed to learn the spatial and temporal features of dMRI data and reconstruct it more efficiently. The extensive quantitative simulations reveal the improvement induced by the proposed method in various settings compared to state-of-the-art methods in dMRI. Further, it considerably advances reconstruction speed from trained dictionaries over the state-of-the-art, permitting faster scans catering to a larger patient group.

Understanding aliasing effects and their removal in SPEN MRI: A k-space perspective.

To characterize the mechanism of formation and the removal of aliasing artifacts and edge ghosts in spatiotemporally encoded (SPEN) MRI within a k-space theoretical framework. SPEN's quadratic phase modulation can be described in k-space by a convolution matrix whose coefficients derive from Fourier relations. This k-space model allows us to pose SPEN's reconstruction as a deconvolution process from which aliasing and edge ghost artifacts can be quantified by estimating the difference between a full sampling and reconstructions resulting from undersampled SPEN data. Aliasing artifacts in SPEN MRI reconstructions can be traced to image contributions corresponding to high-frequency k-space signals. The k-space picture provides the spatial displacements, phase offsets, and linear amplitude modulations associated to these artifacts, as well as routes to removing these from the reconstruction results. These new ways to estimate the artifact priors were applied to reduce SPEN reconstruction artifacts on simulated, phantom, and human brain MRI data. A k-space description of SPEN's reconstruction helps to better understand the signal characteristics of this MRI technique, and to improve the quality of its resulting images.