Image Reconstruction Procedure Research Articles

Super-Resolution Image Reconstruction of Wavefront Coding Imaging System Based on Deep Learning Network

Wavefront Coding (WFC) is an innovative technique aimed at extending the depth of focus (DOF) of optics imaging systems. In digital imaging systems, super-resolution digital reconstruction close to the diffraction limit of optical systems has always been a hot research topic. With the design of a point spread function (PSF) generated by a suitably phase mask, WFC could also be used in super-resolution image reconstruction. In this paper, we use a deep learning network combined with WFC as a general framework for images reconstruction, and verify its possibility and effectiveness. Considering the blur and additive noise simultaneously, we proposed three super-resolution image reconstruction procedures utilizing convolutional neural networks (CNN) based on mean square error (MSE) loss, conditional Generative Adversarial Networks (CGAN), and Swin Transformer Networks (SwinIR) based on mean absolute error (MAE) loss. We verified their effectiveness by simulation experiments. A comparison of experimental results shows that the SwinIR deep residual network structure based on MAE loss optimization criteria can generate more realistic super-resolution images with more details. In addition, we used a WFC camera to obtain a resolution test target and real scene images for experiments. Using the resolution test target, we demonstrated that the spatial resolution could be improved from 55.6 lp/mm to 124 lp/mm by the proposed super-resolution reconstruction procedure. The reconstruction results show that the proposed deep learning network model is superior to the traditional method in reconstructing high-frequency details and effectively suppressing noise, with the resolution approaching the diffraction limit.

Frame rate improvement in ultrafast coherent plane wave compounding

Coherent plane wave compounding (CPWC), as an ultrafast ultrasound imaging technique, makes a significant breakthrough in frame rate enhancement. However, there exists a compromise between the quality of the final image and the frame rate in CPWC. In this paper, we propose an efficient method to minimize the number of required emissions, and consequently, improve the frame rate, while maintaining the image quality. To this end, we down-sample the angle interval using two specific sampling factors. More precisely, we construct two different subsets, each of which consists of a few numbers of emissions. The optimal values of the angle intervals are achieved based on the beampattern that corresponds to the reference case (that is, the case where all plane waves are used). Finally, in order to keep the image quality comparable with the reference case, we apply some modifications to the image reconstruction procedure. In the proposed algorithm, the Delay-and-Sum beamformed images of two considered subsets are convolved to achieve the final reconstructed image. The obtained results confirm the efficiency of the proposed method in terms of frame rate improvement compared to the reference case. In particular, by using the proposed method, the required emissions in PICMUS data reduce to 16, which is 4.6 times smaller compared to the reference case. Also, the gCNR values of the proposed method and the reference case are obtained as 0.98 and 0.97, respectively, for in-vivo dataset. This demonstrates that the proposed method successfully preserves the quality of the reconstructed image by using much fewer emissions.

From Signal to Knowledge: The Diagnostic Value of Raw Data in the Artificial Intelligence Prediction of Human Data for the First Time

Encouraging and astonishing developments have recently been achieved in image-based diagnostic technology. Modern medical care and imaging technology are becoming increasingly inseparable. However, the current diagnosis pattern of signal to image to knowledge inevitably leads to information distortion and noise introduction in the procedure of image reconstruction (from signal to image). Artificial intelligence (AI) technologies that can mine knowledge from vast amounts of data offer opportunities to disrupt established workflows. In this prospective study, for the first time, we develop an AI-based signal-to-knowledge diagnostic scheme for lung nodule classification directly from the computed tomography (CT) raw data (the signal). We find that the raw data achieves almost comparable performance with CT, indicating that it is possible to diagnose diseases without reconstructing images. Moreover, the incorporation of raw data through three common convolutional network structures greatly improves the performance of the CT models in all cohorts (with a gain ranging from 0.01 to 0.12), demonstrating that raw data contains diagnostic information that CT does not possess. Our results break new ground and demonstrate the potential for direct signal-to-knowledge domain analysis.

Spatiotemporal image reconstruction for photoacoustic computed tomography of tumor vascular perfusion in murine models

Monitoring critical physiological processes in murine models like tumor vascular perfusion and its response to prospective anti-cancer treatments is a significant potential use case of dynamic photoacoustic computed tomography (PACT). Previously reported studies of dynamic PACT are based on a frame-by-frame image reconstruction (FBFIR) procedure in which full-view measurement data are assumed to be rapidly acquired. However, many commercial three-dimensional PACT imagers acquire measurements at each tomographic view rotating the object in discrete steps. The time to collect the full-view data is limited by the rotation speed of the object holder and the laser repetition rate. Therefore, FBFIR techniques are not applicable, and there is a critical need for accurate and efficient spatiotemporal image reconstruction (STIR) techniques that can account for spatiotemporal redundancies in the object’s features. To address this, we propose a low-rank matrix estimation-based STIR technique in which the sought-after dynamic image is approximated using a semiseparable approximation in space and time. To validate the proposed method, we also develop a virtual imaging framework for PACT employing dynamic numerical mouse phantoms with a physiologically realistic respiratory motion and perfusion model. The studies demonstrated that the proposed method accurately recovers the tumor vascular perfusion and object motion.

MIMO Coded Generalized Reduced Dimension Fourier Algorithm for 3-D Microwave Imaging

In this paper, to accelerate data acquisition and image reconstruction procedures in a multistatic short-range microwave imaging scenario, an orthogonal coding approach with Fourier domain processing is presented. First, a special two-dimensional (2D) multiple-input multiple-output (MIMO) structure is introduced to fully electronically synthesize the 2D aperture. Then, the model of the transmitted and received signals by a MIMO stepped-frequency-modulated radar is presented, with special considerations about orthogonal, balanced and optimal sequences. On the receiver side, the backscatter frequency response extraction process is formulated with the aim of obtaining individual information of all channels. Finally, based on the introduced model, a fast Fourier-based algorithm with reduced dimensions, named MIMO coded generalized reduced dimension Fourier (CGRDF), is mathematically derived. It includes extracting phase and amplitude compensators with the aim of mapping 4D to 2D spatial data, transferring the backscatter transfer function from the spatial domain to the wavenumber domain, extracting the smoothing filter, compensating the curvature of the wavefront of all scatterers, extracting the reflectivity function and an additional range compensator. The results of numerical simulations show the satisfactory and reliable performance of the proposed approach in terms of the information retrieval process and processing speed.

Performance Improvement of Image-Reconstruction-Based Defense against Adversarial Attack

Deep Neural Networks (DNNs) used for image classification are vulnerable to adversarial examples, which are images that are intentionally generated to predict an incorrect output for a deep learning model. Various defense methods have been proposed to defend against such adversarial attacks, among which, image-reconstruction-based defense methods, such as DIPDefend, are known to be effective in getting rid of the adversarial perturbations injected in the image. However, this image-reconstruction-based defense approach suffers from a long execution time due to its iterative and time-consuming image reconstruction. The trade-off between the execution time and the robustness/accuracy of the defense method should be carefully explored, which is the main focus of this paper. In this work, we aim to improve the execution time of the existing state-of-the-art image-reconstruction-based defense method, DIPDefend, against the Fast Gradient Sign Method (FGSM). In doing so, we propose to take the input-specific properties into consideration when deciding the stopping point of the image reconstruction of DIPDefend. For that, we first applied a low-pass filter to the input image with various kernel sizes to make a prediction of the true label. Then, based on that, the parameters of the image reconstruction procedure were adaptively chosen. Experiments with 500 randomly chosen ImageNet validation set images show that we can obtain an approximately 40% improvement in execution time while keeping the accuracy drop as small as 0.4–3.9%.

A genetic algorithm for image reconstruction in electrical impedance tomography for gesture recognition

Abstract Electrical impedance tomography (EIT) is an imaging method for characterizing the inner conductivity distribution of an object based on the measured boundary voltages resulting from the injection of an AC signal, followed by an image reconstruction procedure. An algorithm tries to solve an ill-posed inverse problem making it challenging to reconstruct an accurate image. To overcome this, we propose a genetic algorithm (GA) for the image reconstruction with a non-blind search method considering prior knowledge about the possible conductivity distribution in the initial search space. To validate the algorithm, experiments have been conducted in a water tank. The algorithm’s performance was evaluated regarding image quality and processing time, being able to minimize the corresponding quality function to 0.0505 with 100 generations using the non-blind search and the uniform crossover/random mutation. Compared to traditional methods, the GA achieves significantly better image quality. It has been implemented as an image reconstruction algorithm for gesture recognition. EIT measurements have been conducted with six persons performing American sign numbers (0–9) resulting in 1800 reconstructed images. They were classified by a previously developed convolutional neural network (CNN), reaching a 92 % accuracy, which is a very good achievement in the case of multiple subjects.

Application of a 2D frequency encoding sectoral approach to hyperpolarized 129Xe MRI at low field

Inhaled hyperpolarized 129Xe MRI is a non-invasive and radiation risk free lung imaging method, which can directly measure the business unit of the lung where gas exchange occurs: the alveoli and acinar ducts (lung function). Currently, three imaging approaches have been demonstrated to be useful for hyperpolarized 129Xe MR in lungs: Fast Gradient Recalled Echo (FGRE), Radial Projection Reconstruction (PR), and spiral/cones. Typically, non-Cartesian acquisitions such as PR and spiral/cones require specific data post-processing, such as interpolating, regridding, and density-weighting procedures for image reconstruction, which often leads to smoothing effects and resolution degradation. On the other hand, Cartesian methods such as FGRE are not short-echo time (TE) methods; they suffer from imaging gradient-induced diffusion-weighting of the k-space center, and employ a significant number of radio-frequency (RF) pulses.Due to the non-renewable magnetization of the hyperpolarized media, the use of a large number of RF pulses (FGRE/PR) required for full k-space coverage is a significant limitation, especially for low field (<0.5 T) hyperpolarized gas MRI.We demonstrate an ultra-fast, purely frequency-encoded, Cartesian pulse sequence called Frequency-Encoding Sectoral (FES), which takes advantage of the long T2* of hyperpolarized 129Xe gas at low field strength (0.074 T). In contrast to PR/FGRE, it uses a much smaller number of RF pulses, and consequently maximizes image Signal-to-Noise Ratio (SNR) while shortening acquisition time. Additionally, FES does not suffer from non-uniform T2* decay leading to image blurring; a common issue with interleaved spirals/cones. The Cartesian k-space coverage of the proposed FES method does not require specific k-space data post-processing, unlike PR/FGRE and spiral/cones methods.Proton scans were used to compare the FES sequence to both FGRE and Phase Encoding Sectoral, in terms of their SNR values and imaging efficiency estimates. Using FES, proton and hyperpolarized 129Xe images were acquired from a custom hollow acrylic phantom (0.04L) and two normal rats (129Xe only), utilizing both single-breath and multiple-breath schemes. For the 129Xe phantom images, the apparent diffusion coefficient, T1, and T2* relaxation maps were acquired and generated. Blurring due to the T2* decay and B0 field variation were simulated to estimate dependence of the image resolution on the duration of the data acquisition windows (i.e. sector length), and temperature-induced resonance frequency shift from the low field magnet hardware.

Quantitative and Anatomical Imaging of Human Skin by Noninvasive Photoacoustic Dermoscopy.

Imaging plays a vital role in the diagnosis and treatment of skin diseases. However, pure optical imaging technique is limited to the visualization of superficial skin tissues. Ultrasonic imaging technique can detect deep tissues, but it lacks detailed information on microscopic pathological structures. Photoacoustic imaging is an advanced technology that bridges the spatial-resolution gap between optical and ultrasonic techniques, by the modes of optical excitation and acoustic detection. Photoacoustic dermoscopy (PAD), based on photoacoustic technology, can noninvasively obtain high-resolution anatomical structures by endogenous absorbers, such as melanin, hemoglobin, lipids, etc. In the past years, PAD has gradually been developed in clinical dermatology for the diagnosis of melanoma, psoriasis, port-wine stains, dermatitis, skin grafting, and testing the efficacy of cosmetics. This protocol provides detailed procedures for PAD construction, including component selection, equipment setup, and system calibration. A step-by-step guide for human skin imaging is provided as an example application. Image reconstruction and troubleshooting procedures are also elaborated. PAD offers the 3D volumetric images of human skin, and quantitatively analyzes the vascular morphology in the dermis. The protocol will provide clinicians with standardized and reasonable guidance in dermatological imaging.

3-D Stochastic Numerical Breast Phantoms for Enabling Virtual Imaging Trials of Ultrasound Computed Tomography.

Ultrasound computed tomography (USCT) is an emerging imaging modality for breast imaging that can produce quantitative images that depict the acoustic properties of tissues. Computer-simulation studies, also known as virtual imaging trials, provide researchers with an economical and convenient route to systematically explore imaging system designs and image reconstruction methods. When simulating an imaging technology intended for clinical use, it is essential to employ realistic numerical phantoms that can facilitate the objective, or task-based, assessment of image quality (IQ). Moreover, when computing objective IQ measures, an ensemble of such phantoms should be employed, which displays the variability in anatomy and object properties that are representative of the to-be-imaged patient cohort. Such stochastic phantoms for clinically relevant applications of USCT are currently lacking. In this work, a methodology for producing realistic 3-D numerical breast phantoms for enabling clinically relevant computer-simulationstudies of USCT breast imaging is presented. By extending and adapting an existing stochastic 3-D breast phantom for use with USCT, methods for creating ensembles of numerical acoustic breast phantoms are established. These breast phantoms will possess clinically relevant variations in breast size, composition, acoustic properties, tumor locations, and tissue textures. To demonstrate the use of the phantoms in virtual USCT studies, two brief case studies are presented, which addresses the development and assessment of image reconstruction procedures. Examples of breast phantoms produced by use of the proposed methods and a collection of 52 sets of simulated USCT measurement data have been made open source for use in image reconstruction development.

An application of filtered back projection method for computed tomography images

AbstractMathematical methods take an important part in reconstruction technologies of radiographic image. Back projection, iterative, and analytical (Two-dimensional Fourier, Filtered Back Projection) methods are the most important procedures for image reconstruction. Whenever there exists numerous projections, analytical methods have a great performance in speed and accuracy and due to these advantages they are comprehensively used for X-ray imaging. One of the widespread used methods in tomographic image reconstruction is Filtered Back Projection (FBP) algorithm. This paper presents an application of this reconstruction algorithm for a generated image of the object. Shepp-Logan filter is used to form the filtered back projection image and performance improvement is investigated. The obtained images indicate that FBP algorithm can be substantial for various applications in the field of medicine and industry.

Transcending conventional snapshot polarimeter performance via neuromorphically adaptive filters.

A channeled Stokes polarimeter that recovers polarimetric signatures across the scene from the modulation induced channels is preferrable for many polarimetric sensing applications. Conventional channeled systems that isolate the intended channels with low-pass filters are sensitive to channel crosstalk effects, and the filters have to be optimized based on the bandwidth profile of scene of interest before applying to each particular scenes to be measured. Here, we introduce a machine learning based channel filtering framework for channeled polarimeters. The machines are trained to predict anti-aliasing filters according to the distribution of the measured data adaptively. A conventional snapshot Stokes polarimeter is simulated to present our machine learning based channel filtering framework. Finally, we demonstrate the advantage of our filtering framework through the comparison of reconstructed polarimetric images with the conventional image reconstruction procedure.

Snapshot photoacoustic topography through an ergodic relay of optical absorption in vivo.

Photoacoustic tomography (PAT) has demonstrated versatile biomedical applications, ranging from tracking single cells to monitoring whole-body dynamics of small animals and diagnosing human breast cancer. Currently, PAT has two major implementations: photoacoustic computed tomography (PACT) and photoacoustic microscopy (PAM). PACT uses a multi-element ultrasonic array for parallel detection, which is relatively complex and expensive. In contrast, PAM requires point-by-point scanning with a single-element detector, which has a limited imaging throughput. The trade-off between the system cost and throughput demands a new imaging method. To this end, we have developed photoacoustic topography through an ergodic relay (PATER). PATER can capture a wide-field image with only a single-element ultrasonic detector upon a single laser shot. This protocol describes the detailed procedures for PATER system construction, including component selection, equipment setup and system alignment. A step-by-step guide for in vivo imaging of a mouse brain is provided as an example application. Data acquisition, image reconstruction and troubleshooting procedures are also elaborated. It takes ~130 min to carry out this protocol, including ~60 min for both calibration and snapshot wide-field data acquisition using a laser with a 2-kHz pulse repetition rate. PATER offers low-cost snapshot wide-field imaging of fast dynamics, such as visualizing blood pulse wave propagation and tracking melanoma tumor cell circulation in mice in vivo. We envision that PATER will have wide biomedical applications and anticipate that the compact size of the setup will allow it to be further developed as a wearable device to monitor human vital signs.

CT iterative vs deep learning reconstruction: comparison of noise and sharpness.

To compare image noise and sharpness of vessels, liver, and muscle in lower extremity CT angiography between "adaptive statistical iterative reconstruction-V" (ASIR-V) and deep learning reconstruction "TrueFidelity" (TFI). Thirty-seven patients (mean age, 65.2 years; 32 men) with lower extremity CT angiography were enrolled between November and December 2019. Images were reconstructed with two ASIR-V (blending factor of 80% and 100% (AV-100)) and three TFI (low-, medium-, and high-strength-level (TF-H) settings). Two radiologists evaluated these images for vessels (aorta, femoral artery, and popliteal artery), liver, and psoas muscle. For quantitative analyses, conventional indicators (CT number, image noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR)) and blur metric values (indicating the degree of image sharpness) of selected regions of interest were determined. For qualitative analyses, the degrees of quantum mottle and blurring were assessed. The higher the blending factor in ASIR-V or the strength in TFI, the lower the noise, the higher the SNR and CNR values, and the higher the blur metric values in all structures. The SNR and CNR values of TF-H images were significantly higher than those of AV-80 images and similar to those of AV-100 images. The blur metric values in TFI images were significantly lower than those in ASIR-V images (p < 0.001), indicating increased sharpness. Among all the investigated image procedures, the overall qualitative image quality was best in TF-H images. TF-H was the most balanced image in terms of image noise and sharpness among the examined image combinations. • Deep learning image reconstruction "TrueFidelity" is superior to iterative reconstruction "ASIR-V" regarding image noise and sharpness. • The high-strength "TrueFidelity" approach generated the best image quality among the examined image reconstruction procedures. • In iterative and deep learning CT image reconstruction, the higher the blending and strength factors, the lower the image noise and the poorer the image sharpness.

Pilot study on deformed bullet correlation

Most studies on bullet identification address test fired bullets that have near pristine striated marks on the land engraved areas (LEAs). However, in case work, bullets found at a crime scene may be severely deformed or fragmented. The resulting missing, expanded, or distorted LEA striations can cause challenges in toolmark comparisons performed by examiners or algorithms. In this paper, an image reconstruction procedure is proposed that, in combination with the Congruent Matching Profile Segments (CMPS) profile comparison method, facilitates the algorithmic correlation of deformed bullets. Initial validation tests were conducted using 57 bullets, with varying degrees of fragmentation or deformation, that were fired from the same 9mm Luger caliber Luger pistol. The bullets spanned 7 different ammunition brands. The CMPS method was applied to correlate the LEA striation profiles extracted from LEA topography images that were corrected for pattern distortion. 15 bullet LEAs, out of 250 bullet LEAs that could be measured, had major distortions. Two sets of comparison tests were conducted, corresponding to a same source and specific source scenario: 1) comparison of the severely distorted LEAs with a near-pristine reference bullet, before and after image reconstruction, and 2) inter comparisons of distorted LEAs, before and after reconstruction. The reconstruction process significantly improved the correlation results when dealing with distorted bullet LEAs. In general, the improvement was larger for samples with relatively large deformation and good striation visibility. Samples with approximately parallel striations tend to have less improvement of CMPS results after profile reconstruction since the CMPS method itself can correct certain scale errors.

Joint image reconstruction and segmentation: Comparison of two algorithms for few-view tomography

Two algorithms of few-view tomography are compared, specifically, the iterative Potts minimization algorithm (IPMA) and the algebraic reconstruction technique with TV-regularization and adaptive segmentation (ART-TVS). Both aim to reconstruct piecewise-constant structures, use the compressed sensing theory, and combine image reconstruction and segmentation procedures. Using a numerical experiment, it is shown that either algorithm can exactly reconstruct the Shepp-Logan phantom from as small as 7 views with noise characteristic of the medical applications of X-ray tomography. However, if an object has a complicated high-frequency structure (QR-code), the minimal number of views required for its exact reconstruction increases to 17–21 for ART-TVS and to 32–34 for IPMA. The ART-TVS algorithm developed by the authors is shown to outperform IPMA in reconstruction accuracy and speed and in resistance to abnormally high noise as well. ART-TVS holds good potential for further improvement.

Dual-image-based reversible data hiding scheme with integrity verification using exploiting modification direction

In this paper, a novel dual-image-based reversible data hiding scheme using exploiting modification direction (EMD) is proposed. This scheme embeds two 5-base secret digits into each pixel pair of the cover image simultaneously according to the EMD matrix to generate two stego-pixel pairs. By shifting these stego-pixel pairs to the appropriate locations in some cases, two meaningful shadows are produced. The secret data can be extracted accurately, and the cover image can be reconstructed completely in the data extraction and the image reconstruction procedure, respectively. Experimental results show that our scheme outperforms the comparative methods in terms of image quality and embedding ratio. Pixel-value differencing (PVD) histogram analysis reveals that our scheme achieves a higher security. Moreover, a tampered shadow can effectively be detected through the proposed scheme.

Computation of exact g-factor maps in 3D GRAPPA reconstructions.

To characterize the noise distributions in 3D-MRI accelerated acquisitions reconstructed with GRAPPA using an exact noise propagation analysis that operates directly in k-space. We exploit the extensive symmetries and separability in the reconstruction steps to account for the correlation between all the acquired k-space samples. Monte Carlo simulations and multi-repetition phantom experiments were conducted to test both the accuracy and feasibility of the proposed method; a high-resolution in-vivo experiment was performed to assess the applicability of our method to clinical scenarios. Our theoretical derivation shows that the direct k-space analysis renders an exact noise characterization under the assumptions of stationarity and uncorrelation in the original k-space. Simulations and phantom experiments provide empirical support to the theoretical proof. Finally, the high-resolution in-vivo experiment demonstrates the ability of the proposed method to assess the impact of the sub-sampling pattern on the overall noise behavior. By operating directly in the k-space, the proposed method is able to provide an exact characterization of noise for any Cartesian pattern sub-sampled along the two phase-encoding directions. Exploitation of the symmetries and separability into independent blocks through the image reconstruction procedure allows us to overcome the computational challenges related to the very large size of the covariance matrices involved.

Visual tracking using structural local DCT sparse appearance model with occlusion detection

In this paper, a structural local DCT sparse appearance model with occlusion detection is proposed for visual tracking in a particle filter framework. The energy compaction property of the 2D-DCT is exploited to reduce the size of the dictionary as well as that of the candidate samples so that the computational cost of l1-minimization can be lowered. Further, a holistic image reconstruction procedure is proposed for robust occlusion detection and used for appearance model update, thus avoiding the degradation of the appearance model in the presence of occlusion/outliers. Also, a patch occlusion ratio is introduced in the confidence score computation to enhance the tracking performance. Quantitative and qualitative performance evaluations on two popular benchmark datasets demonstrate that the proposed tracking algorithm generally outperforms several state-of-the-art methods.

Progressive [formula omitted] secret image sharing schemes based on Boolean operations and covering codes

(k,n) progressive secret image sharing (PSIS) schemes were discussed frequently in recent years. A (k,n) PSIS scheme consists of sharing phase and reconstruction phase. During sharing phase, a secret image is encrypted into n image-shadows; while during reconstruction phase, k to n image-shadows can progressively recover the image; fewer than k image-shadows get no information on the image. Most previous PSIS schemes can be divided into two categories, polynomial based PSIS schemes and visual cryptography based PSIS schemes. Polynomial based schemes can recover high quality image with complicated cryptographic computations; visual cryptography based PSIS schemes can reconstruct image using human visual system without any cryptographic computation, however, the size of image shadow expands largely from the image and the recovered image is quality-distortion. The motivation of this work is to solving the problems in existing PSIS schemes. Thus, in this paper, we adopt three Boolean operations: bit-level XOR, COV(1,7,3) from (7,4) Hamming code and COV(2,8,4) from (8,4) shortened Hamming code to propose three (k,n) PSIS schemes. In our schemes, k to n image-shadows can decrypt randomly partial pixels on entire image, and reconstruct image progressively. Comparing with previous PSIS schemes, our schemes has smaller image-shadow size and higher quality of recovered image than visual cryptography based PSIS schemes. On the other hand, our schemes has much more efficient operations in sharing procedure and image reconstruction procedure than polynomial based PSIS schemes.