Sparse Image Reconstruction Research Articles

Multitask learning-powered large-volume, rapid photoacoustic microscopy with non-diffracting beams excitation and sparse sampling

Large-volume photoacoustic microscopy (PAM) or rapid PAM has attracted increasing attention in biomedical applications due to its ability to provide detailed structural and functional information on tumor pathophysiology and the neuroimmune microenvironment. Non-diffracting beams, such as Airy beams, offer extended depth-of-field (DoF), while sparse image reconstruction using deep learning enables image recovery for rapid imaging. However, Airy beams often introduce side-lobe artifacts, and achieving both extended DoF and rapid imaging remains a challenge, hindering PAM’s adoption as a routine large-volume and repeatable monitoring tool. To address these challenges, we developed multitask learning-powered large-volume, rapid photoacoustic microscopy with Airy beams (ML-LR-PAM). This approach integrates advanced software and hardware solutions designed to mitigate side-lobe artifacts and achieve super-resolution reconstruction. Unlike previous methods that neglect the simultaneous optimization of these aspects, our approach bridges this gap by employing scaled dot-product attention mechanism (SDAM) Wasserstein-based CycleGAN (SW-CycleGAN) for artifact reduction and high-resolution, large-volume imaging. We anticipate that ML-LR-PAM, through this integration, will become a standard tool in both biomedical research and clinical practice.

A novel RCACycleGAN model is proposed for the high-precision reconstruction of sparse TFM images

The sparse total focusing method (TFM) has been shown to enhance the computational efficacy of ultrasound imaging but the image quality of ultrasound regrettably deteriorates with an increase in the sparsity rate of array elements. Deep learning has made remarkable advancements in image processing and cycle-consistent generative adversarial networks (CycleGANs) have been extensively employed to reconstruct diverse image categories. However, due to the incomplete extraction of image feature information by the generator and discriminator in a CycleGAN, high-quality sparse TFM images cannot be directly reconstructed using CycleGANs. There is also a risk of losing crucial feature information related to minor defects. As a result, this paper modifies the generator and discriminator in the CycleGAN to construct a new relativistic discriminator and coordinate attention CycleGAN (RCACycleGAN) model, which enables high-precision reconstruction of sparse TFM images. The addition of the coordinate attention module to the CycleGAN enhances the defective feature representation by fully considering the channel and spatial correlation between regions and using the fusion of spatially perceived feature maps in different directions. It solves the problem of easy loss of defective key feature information. The relativistic discriminator replaces the PatchGAN discriminator in the CycleGAN and evaluates the quality of both real and sparse TFM reconstructed images to ensure a relative image quality evaluation. This process solves the problem of unstable image quality of the sparse TFM reconstructed image. Experimental results demonstrate that RCACycleGAN can stably reconstruct sparse TFM images even in small sample dataset scenarios. The proposed network model reconstructs images with better accuracy, including in terms of structural similarity, defect roundness and area, and has a shorter training time than several existing network models.

UPAMNet: A unified network with deep knowledge priors for photoacoustic microscopy

Photoacoustic microscopy (PAM) has gained increasing popularity in biomedical imaging, providing new opportunities for tissue monitoring and characterization. With the development of deep learning techniques, convolutional neural networks have been used for PAM image resolution enhancement and denoising. However, there exist several inherent challenges for this approach. This work presents a Unified PhotoAcoustic Microscopy image reconstruction Network (UPAMNet) for both PAM image super-resolution and denoising. The proposed method takes advantage of deep image priors by incorporating three effective attention-based modules and a mixed training constraint at both pixel and perception levels. The generalization ability of the model is evaluated in details and experimental results on different PAM datasets demonstrate the superior performance of the method. Experimental results show improvements of 0.59 dB and 1.37 dB, respectively, for 1/4 and 1/16 sparse image reconstruction, and 3.9 dB for image denoising in peak signal-to-noise ratio.

ECT image reconstruction based on sensitive field expansion and optimization

Electrical capacitance tomography (ECT) is an efficient method for addressing the issue of two-phase flow monitoring. Most current methods result in low image reconstruction accuracy due to soft field issues. This paper propose an ECT image reconstruction method based on sensitive field expansion and optimization, which improve reconstruction efficiency and accuracy. Firstly, a sensitivity field optimization method based on flow pattern identification was proposed. The flow pattern recognition determines the flow pattern to which the input signal belongs and selects the sensitive field for the corresponding flow pattern. Secondly, a sensitive field expansion method based on feature extraction is proposed. The ECT image is modeled as a finite rate of innovation signal, sampled using an exponential reproducing kernel. Feature information is extracted from the input signal for data fusion, and the optimized sensitive field is expanded into a new sensitive field distribution matrix by zero-padding and random reorganization. Then, based on optimized and expanded sensitive fields, a sparse image reconstruction method is proposed. Image reconstruction by constructing comprehensive observation equation to obtain the original permittivity distribution vector of the ECT system using the sparsity of the sensitivity matrix and capacitance signal. Finally, the ECT hardware system and the upper computer measurement and imaging integration software are designed. The experimental results show that the method outperforms other existing algorithms in terms of imaging indexes, has better imaging results.

Compressive Perception Image Reconstruction Technology for Basic Mixed Sparse Basis in Metal Surface Detection

<p>Applying Compressed Sensing (CS) technology to robot vision image transmission, an effective method for image reconstruction in robot imaging is proposed to improve the accuracy of reconstruction. Reconstructing images using a mixed sparse representation of DCT and circularly symmetric contour wave transform, the basic algorithm used is the Smoothed Projection Landweber (SPL) algorithm, which optimizes the coefficients under different sparse transformations by incorporating hard thresholding and binary thresholding methods for different sparse bases during iterations. The experiment shows that compared with single sparse base image reconstruction, the proposed reconstruction method has improved reconstruction accuracy.</p> <p> </p>

A nonlinear total variation based computed tomography (CT) image reconstruction method using gradient reinforcement.

Compressed sensing-based reconstruction algorithms have been proven to be more successful than analytical or iterative methods for sparse computed tomography (CT) imaging by narrowing down the solution set thanks to its ability to seek a sparser solution. Total variation (TV), one of the most popular sparsifiers, exploits spatial continuity of features by restricting variation between two neighboring pixels in each direction as using partial derivatives. When the number of projections is much fewer than the one in conventional CT, which results in much less sampling rate than the minimum required one, TV may not provide satisfactory results. In this study, a new regularizer is proposed which seeks for a sparser solution by reinforcing the gradient of TV and empowering the spatial continuity of features. The experiments are done by using both analitical phantom and real human CT images and the results are compared with conventional, four-directional, and directional TV algorithms by using contrast-to-noise ratio (CNR), signal-to-noise ratio (SNR) and Structural Similarity Index (SSIM) metrics. Both quantitative and visual evaluations show that the proposed method is promising for sparse CT image reconstruction by reducing the background noise while preserving the features and edges.

Fixed-time neural networks with time-invariant and time-varying coefficients for mixed variational inequalities

This paper develops several fixed-time neural networks for solving mixed variational inequalities (MVIs). The proposed networks are highly efficient and with fixed-time convergence. First, based on the conventional forward-backward-forward neural network (FNN) and sliding mode control technique, a time-invariant fixed-time FNN (FxTFNN) is designed. Next, the Euclidean norm of FNN is introduced into FxTFNN to design the modified FxTFNN (MFxTFNN). It is shown that the proposed FxTFNN and MFxTFNN have fixed-time convergence properties and their settling-time functions are independent of the initial values. The proposed FxTFNN and MFxTFNN can be used to solve the Lasso problem and apply sparse signal reconstruction and image reconstruction. In addition, by introducing time-varying coefficients based on FxTFNN and MFxTFNN, time-varying FxTFNN (TFxTFNN) and time-varying MFxTFNN (TMFxTFNN) are developed. Finally, experimental results of numerical example, signal reconstruction, and image reconstruction are used to verify the effectiveness and superiority of the proposed neural networks.

Volumetric MRI with sparse sampling for MR-guided 3D motion tracking via sparse prior-augmented implicit neural representation learning.

Volumetric reconstruction of magnetic resonance imaging (MRI) from sparse samples is desirable for 3D motion tracking and promises to improve magnetic resonance (MR)-guided radiation treatment precision. Data-driven sparse MRI reconstruction, however, requires large-scale training datasets for prior learning, which is time-consuming and challenging to acquire in clinical settings. To investigate volumetric reconstruction of MRI from sparse samples of two orthogonal slices aided by sparse priors of two static 3D MRI through implicit neural representation (NeRP) learning, in support of 3D motion tracking during MR-guided radiotherapy. A multi-layer perceptron network was trained to parameterize the NeRP model of a patient-specific MRI dataset, where the network takes 4D data coordinates of voxel locations and motion states as inputs and outputs corresponding voxel intensities. By first training the network to learn the NeRP of two static 3D MRI with different breathing motion states, prior information of patient breathing motion was embedded into network weights through optimization. The prior information was then augmented from two motion states to 31 motion states by querying the optimized network at interpolated and extrapolated motion state coordinates. Starting from the prior-augmented NeRP model as an initialization point, we further trained the network to fit sparse samples of two orthogonal MRI slices and the final volumetric reconstruction was obtained by querying the trained network at 3D spatial locations. We evaluated the proposed method using 5-min volumetric MRI time series with 340ms temporal resolution for seven abdominal patients with hepatocellular carcinoma, acquired using golden-angle radial MRI sequence and reconstructed through retrospective sorting. Two volumetric MRI with inhale and exhale states respectively were selected from the first 30s of the time series for prior embedding and augmentation. The remaining 4.5-min time series was used for volumetric reconstruction evaluation, where we retrospectively subsampled each MRI to two orthogonal slices and compared model-reconstructed images to ground truth images in terms of image quality and the capability of supporting 3D target motion tracking. Across the seven patients evaluated, the peak signal-to-noise-ratio between model-reconstructed and ground truth MR images was 38.02±2.60dB and the structure similarity index measure was 0.98±0.01. Throughout the 4.5-min time period, gross tumor volume (GTV) motion estimated by deforming a reference state MRI to model-reconstructed and ground truth MRI showed good consistency. The 95-percentile Hausdorff distance between GTV contours was 2.41±0.77mm, which is less than the voxel dimension. The mean GTV centroid position difference between ground truth and model estimation was less than 1mm in all three orthogonal directions. A prior-augmented NeRP model has been developed to reconstruct volumetric MRI from sparse samples of orthogonal cine slices. Only one exhale and one inhale 3D MRI were needed to train the model to learn prior information of patient breathing motion for sparse image reconstruction. The proposed model has the potential of supporting 3D motion tracking during MR-guided radiotherapy for improved treatment precision and promises a major simplification of the workflow by eliminating the need for large-scale training datasets.

Radial magnetic resonance image reconstruction with a deep unrolled projected fast iterative soft-thresholding network

Radially sampling of magnetic resonance imaging (MRI) is an effective way to accelerate the imaging. How to preserve the image details in reconstruction is always challenging. In this work, a deep unrolled neural network is designed to emulate the iterative sparse image reconstruction process of a projected fast soft-threshold algorithm (pFISTA). The proposed method, an unrolled pFISTA network for Deep Radial MRI (pFISTA-DR), include the preprocessing module to refine coil sensitivity maps and initial reconstructed image, the learnable convolution filters to extract image feature maps, and adaptive threshold to robustly remove image artifacts. Experimental results show that, among the compared methods, pFISTA-DR provides the best reconstruction and achieved the highest PSNR, the highest SSIM and the lowest reconstruction errors.

High-speed mapping of surface charge dynamics using sparse scanning Kelvin probe force microscopy

Unraveling local dynamic charge processes is vital for progress in diverse fields, from microelectronics to energy storage. This relies on the ability to map charge carrier motion across multiple length- and timescales and understanding how these processes interact with the inherent material heterogeneities. Towards addressing this challenge, we introduce high-speed sparse scanning Kelvin probe force microscopy, which combines sparse scanning and image reconstruction. This approach is shown to enable sub-second imaging (>3 frames per second) of nanoscale charge dynamics, representing several orders of magnitude improvement over traditional Kelvin probe force microscopy imaging rates. Bridging this improved spatiotemporal resolution with macroscale device measurements, we successfully visualize electrochemically mediated diffusion of mobile surface ions on a LaAlO3/SrTiO3 planar device. Such processes are known to impact band-alignment and charge-transfer dynamics at these heterointerfaces. Furthermore, we monitor the diffusion of oxygen vacancies at the single grain level in polycrystalline TiO2. Through temperature-dependent measurements, we identify a charge diffusion activation energy of 0.18 eV, in good agreement with previously reported values and confirmed by DFT calculations. Together, these findings highlight the effectiveness and versatility of our method in understanding ionic charge carrier motion in microelectronics or nanoscale material systems.

Fast iterative regularization by reusing data

AbstractDiscrete inverse problems correspond to solving a system of equations in a stable way with respect to noise in the data. A typical approach to select a meaningful solution is to introduce a regularizer. While for most applications the regularizer is convex, in many cases it is neither smooth nor strongly convex. In this paper, we propose and study two new iterative regularization methods, based on a primal-dual algorithm, to regularize inverse problems efficiently. Our analysis, in the noise free case, provides convergence rates for the Lagrangian and the feasibility gap. In the noisy case, it provides stability bounds and early stopping rules with theoretical guarantees. The main novelty of our work is the exploitation of some a priori knowledge about the solution set: we show that the linear equations determined by the data can be used more than once along the iterations. We discuss various approaches to reuse linear equations that are at the same time consistent with our assumptions and flexible in the implementation. Finally, we illustrate our theoretical findings with numerical simulations for robust sparse recovery and image reconstruction. We confirm the efficiency of the proposed regularization approaches, comparing the results with state-of-the-art methods.

Entangled optical quantum imaging method based on adaptive block compressed sampling

In the process of entangled optical quantum imaging, the Digital Micromirror Device (DMD) sampling mode affects imaging efficiency, and in the process of image sparse reconstruction, the use of a fixed sampling matrix affects the quality of image reconstruction. In this circumstance, this paper proposes an entangled optical quantum imaging method based on Adaptive Block Compressed Sampling with Two-Dimensional Information Entropy (ABCS-2DIE), which measures the texture details of the image through the two-dimensional information entropy of the image, adaptively allocates the sampling rate for each image block according to the texture distribution of the image, and designs the corresponding sampling matrix to sparsely sample the reference light. In addition, coincidence counting between the reference and signal light is considered in the reconstruction algorithm to obtain the target image. Experimental results show that compared with the Overall Compressive Sampling (OCS) method, the proposed method greatly shortens the imaging time and significantly improves the imaging quality. Besides, compared with the Fixed Block Compressed Sampling (FBCS) method, the image obtained by the proposed method has obvious improvement in visual effect, and the corresponding Peak Signal-to-Noise Ratio (PSNR) is improved by 2–4 dB with a slight increase in computation complexity. In all, the proposed method can make good use of the texture features of the image, effectively improving the quality of quantum imaging while reducing the imaging time.

A cascade-based dual-domain data correction network for sparse view CT image reconstruction

Computed tomography (CT) provides non-invasive anatomical structures of the human body and is also widely used for clinical diagnosis, but excessive ionizing radiation in X-rays can cause harm to the human body. Therefore, the researchers obtained sparse sinograms reconstructed sparse view CT images (SVCT) by reducing the amount of X-ray projection, thereby reducing the radiological effects caused by radiation. This paper proposes a cascade-based dual-domain data correction network (CDDCN), which can effectively combine the complementary information contained in the sinogram domain and the image domain to reconstruct high-quality CT images from sparse view sinograms. Specifically, several encoder-decoder subnets are cascaded in the sinogram domain to reconstruct artifact-free and noise-free CT images. In the encoder-decoder subnets, spatial-channel domain learning is designed to achieve efficient feature fusion through a group merging structure, providing continuous and elaborate pixel-level features and improving feature extraction efficiency. At the same time, to ensure that the original sinogram data collected can be retained, a sinogram data consistency layer is proposed to ensure the fidelity of the sinogram data. To further maintain the consistency between the reconstructed image and the reference image, a multi-level composite loss function is designed for regularization to compensate for excessive smoothing and distortion of the image caused by pixel loss and preserve image details and texture. Quantitative and qualitative analysis shows that CDDCN achieves competitive results in artifact removal, edge preservation, detail restoration, and visual improvement for sparsely sampled data under different views.

DIDR-Net: a sparse-view CT deep iterative reconstruction network with an independent detail recovery network

Sparse-view CT is an effective method to reduce X-ray radiation dose in clinical CT imaging. However, sparse view image reconstruction is still challenging due to highly undersampled data. To this end, we propose a new deep learning-based reconstruction model of CT images, called DIDR-Net. Unlike existing methods, DIDR-Net employs a dual network structure, including an iterative reconstruction sub-network and a detail recovery sub-network. The iterative reconstruction sub-network expands the FISTA (Fast Iterative Soft Thresholding Algorithm) into a deep network, utilizing learnable nonlinear sparse transform and shrinkage thresholding to improve reconstruction performance. To avoid loss of image details while removing artifacts, we design a detail recovery sub-network. Specifically, this sub-network captures the local details and global information of the initial image through local and global branches, and adaptively fuses the results of these two branches through a fusion module. DIDR-Net generates both the initial reconstruction map and the detail feature map in a parallel manner, and finally fuses the two sides to reconstruct a high-quality CT image. Experimental results on the AAPM public dataset show that DIDR-Net exhibits better performance in both streak artifact removal and detail structure preservation compared to other advanced reconstruction algorithms.

Motion-resolved fat-fraction mapping with whole-heart free-running multiecho GRE and pilot tone.

To develop a free-running 3D radial whole-heart multiecho gradient echo (ME-GRE) framework for cardiac- and respiratory-motion-resolved fat fraction (FF) quantification. (NTE = 8) readouts optimized for water-fat separation and quantification were integrated within a continuous non-electrocardiogram-triggered free-breathing 3D radial GRE acquisition. Motion resolution was achieved with pilot tone (PT) navigation, and the extracted cardiac and respiratory signals were compared to those obtained with self-gating (SG). After extra-dimensional golden-angle radial sparse parallel-based image reconstruction, FF, R2 *, and B0 maps, as well as fat and water images were generated with a maximum-likelihood fitting algorithm. The framework was tested in a fat-water phantom and in 10 healthy volunteers at 1.5 T using NTE = 4 and NTE = 8 echoes. The separated images and maps were compared with a standard free-breathing electrocardiogram (ECG)-triggered acquisition. The method was validated in vivo, and physiological motion was resolved over all collected echoes. Across volunteers, PT provided respiratory and cardiac signals in agreement (r = 0.91 and r = 0.72) with SG of the first echo, and a higher correlation to the ECG (0.1% of missed triggers for PT vs. 5.9% for SG). The framework enabled pericardial fat imaging and quantification throughout the cardiac cycle, revealing a decrease in FF at end-systole by 11.4% ± 3.1% across volunteers (p < 0.0001). Motion-resolved end-diastolic 3D FF maps showed good correlation with ECG-triggered measurements (FF bias of -1.06%). A significant difference in free-running FF measured with NTE = 4 and NTE = 8 was found (p < 0.0001 in sub-cutaneous fat and p < 0.01 in pericardial fat). Free-running fat fraction mapping was validated at 1.5 T, enabling ME-GRE-based fat quantification with NTE = 8 echoes in 6:15 min.

МЕТОДЫ ИНТЕРПОЛЯЦИИ РАЗРЕЖЕННЫХ ИЗОБРАЖЕНИЙ, РАБОТАЮЩИЕ В ЧАСТОТНОЙ ОБЛАСТИ

The work compares the work of sparse image interpolation methods such as: - interpolation method of sequential computation of the Fourier spectrum (IMSCS), the method of projections onto convex sets (projections onto convex sets, POCS), and the method of amplitude iterations (MAI). All calculations necessary for the reconstruction of sparse images are performed only on spatial spectra. As an example, an aerospace digital image is used, which is typical for the tasks of remote sensing of the earth's surface. A high degree of sparseness is modeled (90 percent of the information is missing). The effectiveness of the studied methods was carried out according to several objective criteria.

Dynamic Orthogonal Matching Pursuit for Sparse Data Reconstruction

The orthogonal matching pursuit (OMP) is one of the mainstream algorithms for sparse data reconstruction or approximation. It acts as a driving force for the development of several other greedy methods for sparse data reconstruction, and it also plays a vital role in the development of compressed sensing theory for sparse signal and image reconstruction. In this paper, we propose the so-called dynamic orthogonal matching pursuit (DOMP) and enhanced dynamic orthogonal matching pursuit (EDOMP) algorithms which are more efficient than OMP for sparse data reconstruction from a numerical point of view. We carry out a rigorous analysis to establish the reconstruction error bound for DOMP under the restricted isometry property of the measurement matrix. The main result claims that the reconstruction error via DOMP can be controlled and measured in terms of the number of iterations, sparsity level of data, and the noise level of measurements. Moveover, the finite convergence of DOMP for a class of large-scale compressed sensing problems is also shown.

Adaptive thresholding for sparse image reconstruction

The performance of the class of sparse reconstruction algorithms which is based on the iterative thresholding is highly dependent on a selection of the appropriate threshold value, controlling a trade-off between the algorithm execution time and the solution accuracy. This is why most of the state-of-the-art reconstruction algorithms employ some method of decreasing the threshold value as the solution converges toward the optimal one. To address this problem we propose a data-driven adaptive threshold selection method based on the fast intersection of confidence intervals (FICI) method, with which we have augmented the two-step iterative shrinkage thresholding (TwIST) algorithm. The performance of the proposed algorithm, denoted as the FICI-TwIST algorithm, has been evaluated on a problem of image reconstruction with the missing pixels, exploiting image sparsity in the discrete cosine transformation domain. The obtained results have shown competitive performance in comparison with a number of state-of-the-art sparse reconstruction algorithms, even outperforming them in some scenarios.

De-Aliasing and Accelerated Sparse Magnetic Resonance Image Reconstruction Using Fully Dense CNN with Attention Gates.

When sparsely sampled data are used to accelerate magnetic resonance imaging (MRI), conventional reconstruction approaches produce significant artifacts that obscure the content of the image. To remove aliasing artifacts, we propose an advanced convolutional neural network (CNN) called fully dense attention CNN (FDA-CNN). We updated the Unet model with the fully dense connectivity and attention mechanism for MRI reconstruction. The main benefit of FDA-CNN is that an attention gate in each decoder layer increases the learning process by focusing on the relevant image features and provides a better generalization of the network by reducing irrelevant activations. Moreover, densely interconnected convolutional layers reuse the feature maps and prevent the vanishing gradient problem. Additionally, we also implement a new, proficient under-sampling pattern in the phase direction that takes low and high frequencies from the k-space both randomly and non-randomly. The performance of FDA-CNN was evaluated quantitatively and qualitatively with three different sub-sampling masks and datasets. Compared with five current deep learning-based and two compressed sensing MRI reconstruction techniques, the proposed method performed better as it reconstructed smoother and brighter images. Furthermore, FDA-CNN improved the mean PSNR by 2 dB, SSIM by 0.35, and VIFP by 0.37 compared with Unet for the acceleration factor of 5.

Centralized and Collective Neurodynamic Optimization Approaches for Sparse Signal Reconstruction via L₁-Minimization.

This article develops several centralized and collective neurodynamic approaches for sparse signal reconstruction by solving the L1 -minimization problem. First, two centralized neurodynamic approaches are designed based on the augmented Lagrange method and the Lagrange method with derivative feedback and projection operator. Then, the optimality and global convergence of them are derived. In addition, considering that the collective neurodynamic approaches have the function of information protection and distributed information processing, first, under mild conditions, we transform the L1 -minimization problem into two network optimization problems. Later, two collective neurodynamic approaches based on the above centralized neurodynamic approaches and multiagent consensus theory are proposed to address the obtained network optimization problems. As far as we know, this is the first attempt to use the collective neurodynamic approaches to deal with the L1 -minimization problem in a distributed manner. Finally, several comparative experiments on sparse signal and image reconstruction demonstrate that our proposed centralized and collective neurodynamic approaches are efficient and effective.