Image Reconstruction Framework Research Articles

Breaking through clouds: A hierarchical fusion network empowered by dual-domain cross-modality interactive attention for cloud-free image reconstruction

Cloud obscuration undermines the availability of optical images for continuous monitoring in earth observation. Fusing features from synthetic aperture radar (SAR) has been recognized as a feasible strategy to guide the reconstruction of corrupted signals in cloud-contaminated regions. However, due to the different imaging mechanisms and reflection characteristics, the substantial domain gap between SAR images and optical images makes it a challenging problem to effectively perform cross-modality feature fusion. Although several SAR-assisted cloud removal methods have been proposed, most of them are often unable to achieve adequate information interaction between different modalities, which greatly limits the effectiveness and reliability of cross-modality fusion. In this paper, we proposed a novel hierarchical framework for cloud-free multispectral image reconstruction, which effectively integrates SAR and optical data by a dual-domain interactive attention mechanism. The overall encoder–decoder network is a W-shaped asymmetric structure with a two-branch encoder and one decoder. The encoder branches extract features from SAR images and optical images separately, while the decoder exploits multiscale residual block groups to expand the receptive field and multi-output strategy for reducing the training difficulty. The core cross-modality feature fusion module at the bottleneck adopts the dual-domain interactive attention (DDIA) mechanism which enhances the reciprocal infusion of SAR and optical features to encourage the reconstruction of spectral and structural information. Furthermore, features in the spatial and frequency domains are integrated to improve the effectiveness of the fusion process. To echo the overall network structure, the loss function is designed as a multiscale loss in dual domains. The proposed method can realize sufficient information communication and effective cross-modality fusion between SAR features and optical features. Extensive experiments on the SMILE-CR dataset and SEN12MS-CR dataset demonstrated that the proposed method can outperform the seven representative deep-learning comparative methods in terms of visual performance and quantitative accuracy.

Single-pixel complex-amplitude imaging based on untrained complex-valued convolutional neural network

Single-pixel imaging is advancing rapidly in complex-amplitude imaging. However, reconstructing high-quality images demands significant acquisition and heavy computation, making the entire imaging process time-consuming. Here we propose what we believe to be a novel single-pixel complex-amplitude imaging (SCI) scheme using a complex-valued convolutional neural network for image reconstruction. The proposed sheme does not need to pre-train on any labeled data, and can quickly reconstruct high-quality complex-amplitude images with the randomly initialized network only under the constraints of the physical model. Simulation and experimental results show that the proposed scheme is effective and feasible, and can achieve a good balance between efficiency and quality. We believe that this work provides a new image reconstruction framework for SCI, and paves the way for its practical applications.

Structure-aware diffusion for low-dose CT imaging

Reducing the radiation dose leads to the x-ray computed tomography (CT) images suffering from heavy noise and artifacts, which inevitably interferes with the subsequent clinic diagnostic and analysis. Leading works have explored diffusion models for low-dose CT imaging to avoid the structure degeneration and blurring effects of previous deep denoising models. However, most of them always begin their generative processes with Gaussian noise, which has little or no structure priors of the clean data distribution, thereby leading to long-time inference and unpleasant reconstruction quality. To alleviate these problems, this paper presents a Structure-Aware Diffusion model (SAD), an end-to-end self-guided learning framework for high-fidelity CT image reconstruction. First, SAD builds a nonlinear diffusion bridge between clean and degraded data distributions, which could directly learn the implicit physical degradation prior from observed measurements. Second, SAD integrates the prompt learning mechanism and implicit neural representation into the diffusion process, where rich and diverse structure representations extracted by degraded inputs are exploited as prompts, which provides global and local structure priors, to guide CT image reconstruction. Finally, we devise an efficient self-guided diffusion architecture using an iterative updated strategy, which further refines structural prompts during each generative step to drive finer image reconstruction. Extensive experiments on AAPM-Mayo and LoDoPaB-CT datasets demonstrate that our SAD could achieve superior performance in terms of noise removal, structure preservation, and blind-dose generalization, with few generative steps, even one step only.

Image reconstruction of multispectral sparse sampling photoacoustic tomography based on deep algorithm unrolling

Photoacoustic tomography (PAT), as a novel medical imaging technology, provides structural, functional, and metabolism information of biological tissue in vivo. Sparse Sampling PAT, or SS-PAT, generates images with a smaller number of detectors, yet its image reconstruction is inherently ill-posed. Model-based methods are the state-of-the-art method for SS-PAT image reconstruction, but they require design of complex handcrafted prior. Owing to their ability to derive robust prior from labeled datasets, deep-learning-based methods have achieved great success in solving inverse problems, yet their interpretability is poor. Herein, we propose a novel SS-PAT image reconstruction method based on deep algorithm unrolling (DAU), which integrates the advantages of model-based and deep-learning-based methods. We firstly provide a thorough analysis of DAU for PAT reconstruction. Then, in order to incorporate the structural prior constraint, we propose a nested DAU framework based on plug-and-play Alternating Direction Method of Multipliers (PnP-ADMM) to deal with the sparse sampling problem. Experimental results on numerical simulation, in vivo animal imaging, and multispectral un-mixing demonstrate that the proposed DAU image reconstruction framework outperforms state-of-the-art model-based and deep-learning-based methods.

LMD: Faster Image Reconstruction with Latent Masking Diffusion

As a class of fruitful approaches, diffusion probabilistic models (DPMs) have shown excellent advantages in high-resolution image reconstruction. On the other hand, masked autoencoders (MAEs), as popular self-supervised vision learners, have demonstrated simpler and more effective image reconstruction and transfer capabilities on downstream tasks. However, they all require extremely high training costs, either due to inherent high temporal-dependence (i.e., excessively long diffusion steps) or due to artificially low spatial-dependence (i.e., human-formulated high mask ratio, such as 0.75). To the end, this paper presents LMD, a faster image reconstruction framework with Latent Masking Diffusion. First, we propose to project and reconstruct images in latent space through a pre-trained variational autoencoder, which is theoretically more efficient than in the pixel-based space. Then, we combine the advantages of MAEs and DPMs to design a progressive masking diffusion model, which gradually increases the masking proportion by three different schedulers and reconstructs the latent features from simple to difficult, without sequentially performing denoising diffusion as in DPMs or using fixed high masking ratio as in MAEs, so as to alleviate the high training time-consumption predicament. Our approach allows for learning high-capacity models and accelerate their training (by 3x or more) and barely reduces the original accuracy. Inference speed in downstream tasks also significantly outperforms the previous approaches.

Model-based iterative reconstruction for direct imaging with point spread function encoded echo planar MRI

BackgroundEcho planar imaging (EPI) is a fast measurement technique commonly used in magnetic resonance imaging (MRI), but is highly sensitive to measurement non-idealities in reconstruction. Point spread function (PSF)-encoded EPI is a multi-shot strategy which alleviates distortion, but acquisition of encodings suitable for direct distortion-free imaging prolongs scan time. In this work, a model-based iterative reconstruction (MBIR) framework is introduced for direct imaging with PSF-EPI to improve image quality and acceleration potential. MethodsAn MBIR platform was developed for accelerated PSF-EPI. The reconstruction utilizes a subspace representation, is regularized to promote local low-rankedness (LLR), and uses variable splitting for efficient iteration. Comparisons were made against standard reconstructions from prospectively accelerated PSF-EPI data and with retrospective subsampling. Exploring aggressive partial Fourier acceleration of the PSF-encoding dimension, additional comparisons were made against an extension of Homodyne to direct PSF-EPI in numerical experiments. A neuroradiologists' assessment was completed comparing images reconstructed with MBIR from retrospectively truncated data directly against images obtained with standard reconstructions from non-truncated datasets. ResultsImage quality results were consistently superior for MBIR relative to standard and Homodyne reconstructions. As the MBIR signal model and reconstruction allow for arbitrary sampling of the PSF space, random sampling of the PSF-encoding dimension was also demonstrated, with quantitative assessments indicating best performance achieved through nonuniform PSF sampling combined with partial Fourier. With retrospective subsampling, MBIR reconstructs high-quality images from sub-minute scan datasets. MBIR was shown to be superior in a neuroradiologists' assessment with respect to three of five performance criteria, with equivalence for the remaining two. ConclusionsA novel image reconstruction framework is introduced for direct imaging with PSF-EPI, enabling arbitrary PSF space sampling and reconstruction of diagnostic-quality images from highly accelerated PSF-encoded EPI data.

Development of a Monte Carlo-based scatter correction method for total-body PET using the uEXPLORER PET/CT scanner

Objective. This study presents and evaluates a robust Monte Carlo-based scatter correction (SC) method for long axial field of view (FOV) and total-body positron emission tomography (PET) using the uEXPLORER total-body PET/CT scanner. Approach. Our algorithm utilizes the Monte Carlo (MC) tool SimSET to compute SC factors in between individual image reconstruction iterations within our in-house list-mode and time-of-flight-based image reconstruction framework. We also introduced a unique scatter scaling technique at the detector block-level for optimal estimation of the scatter contribution in each line of response. First image evaluations were derived from phantom data spanning the entire axial FOV along with image data from a human subject with a large body mass index. Data was evaluated based on qualitative inspections, and contrast recovery, background variability, residual scatter removal from cold regions, biases and axial uniformity were quantified and compared to non-scatter-corrected images. Main results. All reconstructed images demonstrated qualitative and quantitative improvements compared to non-scatter-corrected images: contrast recovery coefficients improved by up to 17.2% and background variability was reduced by up to 34.3%, and the residual lung error was between 1.26% and 2.08%. Low biases throughout the axial FOV indicate high quantitative accuracy and axial uniformity of the corrections. Up to 99% of residual activity in cold areas in the human subject was removed, and the reliability of the method was demonstrated in challenging body regions like in the proximity of a highly attenuating knee prosthesis. Significance. The MC SC method employed was demonstrated to be accurate and robust in TB-PET. The results of this study can serve as a benchmark for optimizing the quantitative performance of future SC techniques.

Frequency-aware network for low-light image enhancement

Low-light images often suffer from severe visual degradation, affecting both human perception and high-level computer vision tasks. Most existing methods process images in the spatial domain, making it challenging to simultaneously improve brightness while suppressing noise. In this paper, we present a novel perspective to enhance images based on frequency domain characteristics. Specifically, we reveal that the low-frequency components are closely related to luminance and color, whereas the high-frequency components are not. Based on this observation, we propose the Frequency-aware Network (FaNet) for low-light image enhancement. By selectively adjusting low-frequency components, FaNet preserves more high-frequency details while achieving low-light image enhancement. Additionally, we employ a multi-scale framework and selective fusion for effective feature learning and image reconstruction. Experimental results demonstrate the superiority of the proposed method.

Cross-Domain Unpaired Learning for Low-Dose CT Imaging.

Supervised deep-learning techniques with paired training datasets have been widely studied for low-dose computed tomography (LDCT) imaging with excellent performance. However, the paired training datasets are usually difficult to obtain in clinical routine, which restricts the wide adoption of supervised deep-learning techniques in clinical practices. To address this issue, a general idea is to construct a pseudo paired training dataset based on the widely available unpaired data, after which, supervised deep-learning techniques can be adopted for improving the LDCT imaging performance by training on the pseudo paired training dataset. However, due to the complexity of noise properties in CT imaging, the LDCT data are difficult to generate in order to construct the pseudo paired training dataset. In this paper, we propose a simple yet effective cross-domain unpaired learning framework for pseudo LDCT data generation and LDCT image reconstruction, which is denoted as CrossDuL. Specifically, a dedicated pseudo LDCT sinogram generative module is constructed based on a data-dependent noise model in the sinogram domain, and then instead of in the sinogram domain, a pseudo paired dataset is constructed in the image domain to train an LDCT image restoration module. To validate the effectiveness of the proposed framework, clinical datasets are adopted. Experimental results demonstrate that the CrossDuL framework can obtain promising LDCT imaging performance in both quantitative and qualitative measurements.

Eddy current-induced artifact correction in high b-value ex vivo human brain diffusion MRI with dynamic field monitoring.

To investigate whether spatiotemporal magnetic field monitoring can correct pronounced eddy current-induced artifacts incurred by strong diffusion-sensitizing gradients up to 300 mT/m used in high b-value diffusion-weighted (DW) EPI. A dynamic field camera equipped with 16 1 H NMR field probes was first used to characterize field perturbations caused by residual eddy currents from diffusion gradients waveforms in a 3D multi-shot EPI sequence on a 3T Connectom scanner for different gradient strengths (up to 300 mT/m), diffusion directions, and shots. The efficacy of dynamic field monitoring-based image reconstruction was demonstrated on high-gradient strength, submillimeter resolution whole-brain ex vivo diffusion MRI. A 3D multi-shot image reconstruction framework was developed that incorporated the nonlinear phase evolution measured with the dynamic field camera. Phase perturbations in the readout induced by residual eddy currents from strong diffusion gradients are highly nonlinear in space and time, vary among diffusion directions, and interfere significantly with the image encoding gradients, changing the k-space trajectory. During the readout, phase modulations between odd and even EPI echoes become non-static and diffusion encoding direction-dependent. Superior reduction of ghosting and geometric distortion was achieved with dynamic field monitoring compared to ghosting reduction approaches such as navigator- and structured low-rank-based methods or MUSE followed by image-based distortion correction with the FSL tool "eddy." Strong eddy current artifacts characteristic of high-gradient strength DW-EPI can be well corrected with dynamic field monitoring-based image reconstruction.

Joint Image Reconstruction and Super-Resolution for Accelerated Magnetic Resonance Imaging.

Magnetic resonance (MR) image reconstruction and super-resolution are two prominent techniques to restore high-quality images from undersampled or low-resolution k-space data to accelerate MR imaging. Combining undersampled and low-resolution acquisition can further improve the acceleration factor. Existing methods often treat the techniques of image reconstruction and super-resolution separately or combine them sequentially for image recovery, which can result in error propagation and suboptimal results. In this work, we propose a novel framework for joint image reconstruction and super-resolution, aiming to efficiently image recovery and enable fast imaging. Specifically, we designed a framework with a reconstruction module and a super-resolution module to formulate multi-task learning. The reconstruction module utilizes a model-based optimization approach, ensuring data fidelity with the acquired k-space data. Moreover, a deep spatial feature transform is employed to enhance the information transition between the two modules, facilitating better integration of image reconstruction and super-resolution. Experimental evaluations on two datasets demonstrate that our proposed method can provide superior performance both quantitatively and qualitatively.

Comparison of free-running whole-heart 5D and 4D flow imaging to standard 2D phase contrast in patients with right-sided congenital heart disease

Abstract Funding Acknowledgements Type of funding sources: None. Background Flow measurements by cardiac magnetic resonance (CMR) are part of a routine assessment in patients with congenital heart disease (CHD). Standard 2D phase contrast (2DPC) sequences, however, require precise prescription of image planes leading to long scan times which could be overcome by 3D-acquisitons covering the entire cardiac anatomy in a single acquisition. Thus, recently introduced accelerated free-breathing 3D whole heart flow sequences promise to simplify CMR exams allowing an easier acquisition and retrospective flow measurements in virtually all vessels. Purpose Comparison of standard 2DPC to an accelerated 4D flow sequence with prospective respiratory navigation (1), and a free-running radial fully self-gated respiratory and cardiac motion-resolved 5D flow sequence (5D flow) (2,3). Methods Patients with right-sided CHD and no intra- or extracardiac shunt underwent a CMR on a 1.5T scanner. 2DPC flow sequences and a 4D flow data set were obtained (1). In addition, a whole-heart free-running 3D radial PC CMR sequence was acquired (2). Using a compressed-sensing-based image reconstruction framework (2,3), 5D flow images were obtained, and the end-expiratory phase was used for analysis. 3D vessel segmentation was based on retrospectively computed phase-contrast angiography images for each 4D and 5D flow datasets. Flow measured in the ascending, descending aorta (AA, DA), main, right and left pulmonary artery (MPA, RPA, LPA) and superior vena cava (SVC) as well as the internal consistency for determination of systemic and pulmonary output were compared between all three sequences by Bland-Altman analyses. Results Twenty CHD patients (age 31±11y, 7 women, tetralogy of Fallot N=7, Ross operation N=10; pulmonary regurgitation, atrial switch and Ebstein’s disease each N=1) were included. Mean peak velocity of MPA/ right ventricle to pulmonary artery conduit was 1.9 (range 0.9; 3.5) m/s. Compared to 2DPC, both 4D and 5D flow moderately underestimated absolute flow with significant lower values for the RPA for 4D flow (Table 1). The bias (i.e. the mean absolute difference) of MPA flow between 2DPC and 4D flow was related to the MPA peak velocity (r2=0.3, p = 0.03) with a similar trend for the comparison of 2DPC vs. 5D flow (r2=0.142, p = 0.141). The comparison of the internal consistency for determination of systemic and pulmonary output showed similar results for 2DPC and 5D flow and a slightly lower agreement for the 4D flow sequence (Table 2). Conclusions In this preliminary study, 4D and 5D flow provide results comparable to 2DPC, however, both sequences showed a trend towards lower absolute flow values. Complex flow patterns due to the right-sided pathologies appear to have an impact on the difference observed between sequences as shown for the MPA. The internal consistency of flow quantification was similar for the 2DPC and 5D flow sequences. The presented observations are currently being evaluated in a larger study.

Non-uniform image reconstruction for fast photoacoustic microscopy of histology imaging.

Photoacoustic microscopic imaging utilizes the characteristic optical absorption properties of pigmented materials in tissues to enable label-free observation of fine morphological and structural features. Since DNA/RNA can strongly absorb ultraviolet light, ultraviolet photoacoustic microscopy can highlight the cell nucleus without complicated sample preparations such as staining, which is comparable to the standard pathological images. Further improvements in the imaging acquisition speed are critical to advancing the clinical translation of photoacoustic histology imaging technology. However, improving the imaging speed with additional hardware is hampered by considerable costs and complex design. In this work, considering heavy redundancy in the biological photoacoustic images that overconsume the computing power, we propose an image reconstruction framework called non-uniform image reconstruction (NFSR), which exploits an object detection network to reconstruct low-sampled photoacoustic histology images into high-resolution images. The sampling speed of photoacoustic histology imaging is significantly improved, saving 90% of the time cost. Furthermore, NFSR focuses on the reconstruction of the region of interest while maintaining high PSNR and SSIM evaluation indicators of more than 99% but reducing the overall computation by 60%.

Exploiting volumetric wave correlation for enhanced depth imaging in scattering medium

Imaging an object embedded within a scattering medium requires the correction of complex sample-induced wave distortions. Existing approaches have been designed to resolve them by optimizing signal waves recorded in each 2D image. Here, we present a volumetric image reconstruction framework that merges two fundamental degrees of freedom, the wavelength and propagation angles of light waves, based on the object momentum conservation principle. On this basis, we propose methods for exploiting the correlation of signal waves from volumetric images to better cope with multiple scattering. By constructing experimental systems scanning both wavelength and illumination angle of the light source, we demonstrated a 32-fold increase in the use of signal waves compared with that of existing 2D-based approaches and achieved ultrahigh volumetric resolution (lateral resolution: 0.41 mu m, axial resolution: 0.60 mu m) even within complex scattering medium owing to the optimal coherent use of the broad spectral bandwidth (225 nm).

A Deep Learning Framework for Cardiac MR Under-Sampled Image Reconstruction with a Hybrid Spatial and k-Space Loss Function.

Magnetic resonance imaging (MRI) is an efficient, non-invasive diagnostic imaging tool for a variety of disorders. In modern MRI systems, the scanning procedure is time-consuming, which leads to problems with patient comfort and causes motion artifacts. Accelerated or parallel MRI has the potential to minimize patient stress as well as reduce scanning time and medical costs. In this paper, a new deep learning MR image reconstruction framework is proposed to provide more accurate reconstructed MR images when under-sampled or aliased images are generated. The proposed reconstruction model is designed based on the conditional generative adversarial networks (CGANs) where the generator network is designed in a form of an encoder-decoder U-Net network. A hybrid spatial and k-space loss function is also proposed to improve the reconstructed image quality by minimizing the L1-distance considering both spatial and frequency domains simultaneously. The proposed reconstruction framework is directly compared when CGAN and U-Net are adopted and used individually based on the proposed hybrid loss function against the conventional L1-norm. Finally, the proposed reconstruction framework with the extended loss function is evaluated and compared against the traditional SENSE reconstruction technique using the evaluation metrics of structural similarity (SSIM) and peak signal to noise ratio (PSNR). To fine-tune and evaluate the proposed methodology, the public Multi-Coil k-Space OCMR dataset for cardiovascular MR imaging is used. The proposed framework achieves a better image reconstruction quality compared to SENSE in terms of PSNR by 6.84 and 9.57 when U-Net and CGAN are used, respectively. Similarly, it demonstrates SSIM of the reconstructed MR images comparable to the one provided by the SENSE algorithm when U-Net and CGAN are used. Comparing cases where the proposed hybrid loss function is used against the cases with the simple L1-norm, the reconstruction performance can be noticed to improve by 6.84 and 9.57 for U-Net and CGAN, respectively. To conclude this, the proposed framework using CGAN provides the best reconstruction performance compared with U-Net or the conventional SENSE reconstruction techniques. The proposed framework seems to be useful for the practical reconstruction of cardiac images since it can provide better image quality in terms of SSIM and PSNR.

Deep Learning-Based Reconstruction for Cardiac MRI: A Review

Cardiac magnetic resonance (CMR) is an essential clinical tool for the assessment of cardiovascular disease. Deep learning (DL) has recently revolutionized the field through image reconstruction techniques that allow unprecedented data undersampling rates. These fast acquisitions have the potential to considerably impact the diagnosis and treatment of cardiovascular disease. Herein, we provide a comprehensive review of DL-based reconstruction methods for CMR. We place special emphasis on state-of-the-art unrolled networks, which are heavily based on a conventional image reconstruction framework. We review the main DL-based methods and connect them to the relevant conventional reconstruction theory. Next, we review several methods developed to tackle specific challenges that arise from the characteristics of CMR data. Then, we focus on DL-based methods developed for specific CMR applications, including flow imaging, late gadolinium enhancement, and quantitative tissue characterization. Finally, we discuss the pitfalls and future outlook of DL-based reconstructions in CMR, focusing on the robustness, interpretability, clinical deployment, and potential for new methods.

Multi-channel GAN-based calibration-free diffusion-weighted liver imaging with simultaneous coil sensitivity estimation and reconstruction.

Diffusion-weighted imaging (DWI) with parallel reconstruction may suffer from a mismatch between the coil calibration scan and imaging scan due to motions, especially for abdominal imaging. This study aimed to construct an iterative multichannel generative adversarial network (iMCGAN)-based framework for simultaneous sensitivity map estimation and calibration-free image reconstruction. The study included 106 healthy volunteers and 10 patients with tumors. The performance of iMCGAN was evaluated in healthy participants and patients and compared with the SAKE, ALOHA-net, and DeepcomplexMRI reconstructions. The peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), root mean squared error (RMSE), and histograms of apparent diffusion coefficient (ADC) maps were calculated for assessing image qualities. The proposed iMCGAN outperformed the other methods in terms of the PSNR (iMCGAN: 41.82 ± 2.14; SAKE: 17.38 ± 1.78; ALOHA-net: 20.43 ± 2.11 and DeepcomplexMRI: 39.78 ± 2.78) for b = 800 DWI with an acceleration factor of 4. Besides, the ghosting artifacts in the SENSE due to the mismatch between the DW image and the sensitivity maps were avoided using the iMCGAN model. The current model iteratively refined the sensitivity maps and the reconstructed images without additional acquisitions. Thus, the quality of the reconstructed image was improved, and the aliasing artifact was alleviated when motions occurred during the imaging procedure.

RISING: A new framework for model-based few-view CT image reconstruction with deep learning.

Medical image reconstruction from low-dose tomographic data is an active research field, recently revolutionized by the advent of deep learning. In fact, deep learning typically yields superior results than classical optimization approaches, but unstable results have been reported both numerically and theoretically in the literature. This paper proposes RISING, a new framework for sparse-view tomographic image reconstruction combining an early-stopped Rapid Iterative Solver with a subsequent Iteration Network-based Gaining step. In our two-step approach, the first phase executes very few iterations of a regularized model-based algorithm, whereas the second step completes the missing iterations by means of a convolutional neural network. The proposed method is ground-truth free; it exploits the computational speed and flexibility of a data-driven approach, but it also imposes sparsity constraints to the solution as in the model-based setting. Experiments performed both on a digitally created and on a real abdomen data set confirm the numerical and visual accuracy of the reconstructed RISING images in short computational times. These features make the framework promising to be used on real systems for clinical exams.

A jointed feature fusion framework for photoacoustic image reconstruction

The standard reconstruction of Photoacoustic (PA) computed tomography (PACT) image could cause the artifacts due to interferences or ill-posed setup. Recently, deep learning has been used to reconstruct the PA image with ill-posed conditions. In this paper, we propose a jointed feature fusion framework (JEFF-Net) based on deep learning to reconstruct the PA image using limited-view data. The cross-domain features from limited-view position-wise data and the reconstructed image are fused by a backtracked supervision. A quarter position-wise data (32 channels) is fed into model, which outputs another 3-quarters-view data (96 channels). Moreover, two novel losses are designed to restrain the artifacts by sufficiently manipulating superposed data. The experimental results have demonstrated the superior performance and quantitative evaluations show that our proposed method outperformed the ground-truth in some metrics by 135% (SSIM for simulation) and 40% (gCNR for in-vivo) improvement.

SelfCoLearn: Self-Supervised Collaborative Learning for Accelerating Dynamic MR Imaging.

Lately, deep learning technology has been extensively investigated for accelerating dynamic magnetic resonance (MR) imaging, with encouraging progresses achieved. However, without fully sampled reference data for training, the current approaches may have limited abilities in recovering fine details or structures. To address this challenge, this paper proposes a self-supervised collaborative learning framework (SelfCoLearn) for accurate dynamic MR image reconstruction from undersampled k-space data directly. The proposed SelfCoLearn is equipped with three important components, namely, dual-network collaborative learning, reunderampling data augmentation and a special-designed co-training loss. The framework is flexible and can be integrated into various model-based iterative un-rolled networks. The proposed method has been evaluated on an in vivo dataset and was compared to four state-of-the-art methods. The results show that the proposed method possesses strong capabilities in capturing essential and inherent representations for direct reconstructions from the undersampled k-space data and thus enables high-quality and fast dynamic MR imaging.