Image Reconstruction Methodology Research Articles

Uncertainty Quantification for Deep Unrolling-Based Computational Imaging

Deep unrolling is an emerging deep learning-based image reconstruction methodology that bridges the gap between model-based and purely deep learning-based image reconstruction methods. Although deep unrolling methods achieve state-of-the-art performance for imaging problems and allow the incorporation of the observation model into the reconstruction process, they do not provide any uncertainty information about the reconstructed image, which severely limits their use in practice, especially for safety-critical imaging applications. In this paper, we propose a learning-based image reconstruction framework that incorporates the observation model into the reconstruction task and that is capable of quantifying epistemic and aleatoric uncertainties, based on deep unrolling and Bayesian neural networks. We demonstrate the uncertainty characterization capability of the proposed framework on magnetic resonance imaging and computed tomography reconstruction problems. We investigate the characteristics of the epistemic and aleatoric uncertainty information provided by the proposed framework to motivate future research on utilizing uncertainty information to develop more accurate, robust, trustworthy, uncertainty-aware, learning-based image reconstruction and analysis methods for imaging problems. We show that the proposed framework can provide uncertainty information while achieving comparable reconstruction performance to state-of-the-art deep unrolling methods.

Ultrahigh-Resolution, Label-Free Hyperlens Imaging in the Mid-IR.

The hyperbolic phonon polaritons supported in hexagonal boron nitride (hBN) with long scattering lifetimes are advantageous for applications such as super-resolution imaging via hyperlensing. Yet, hyperlens imaging is challenging for distinguishing individual and closely spaced objects and for correlating the complicated hyperlens fields with the structure of an unknown object underneath. Here, we make significant strides to overcome each of these challenges. First, we demonstrate that monoisotopic h11BN provides significant improvements in spatial resolution, experimentally resolving structures as small as 44 nm and those with sub 25 nm spacings at 6.76 μm free-space wavelength. We also present an image reconstruction algorithm that provides a structurally accurate, visual representation of the embedded objects from the complex hyperlens field. Further, we offer additional insights into optimizing hyperlens performance on the basis of material properties, with an eye toward realizing far-field imaging modalities. Thus, our results significantly advance label-free, high-resolution, spectrally selective hyperlens imaging and image reconstruction methodologies.

Full image reconstruction in frequency-domain photoacoustic microscopy by means of a low-cost I/Q demodulator.

We present a full image reconstruction methodology in frequency-domain photoacoustic (PA) microscopy using a low-cost I/Q demodulator for the recording of the amplitude and phase of the signals. By modulating the intensity of a continuous-wave diode laser at 10 MHz, we have been able to provide accurate optical absorption images and surface reconstructions of phantom samples, comparing also the extracted results with standard time-domain approaches. The findings of the study in this Letter could be utilized towards the development of inexpensive PA microscopes with multispectral capabilities for a wide range of biomedical studies, requiring the sensitive detection of endogenous or exogenous absorbers in tissues.

Factors Affecting Radiation Dose in Computed Tomography Angiograms for Pulmonary Embolism: A Retrospective Cohort Study.

Objectives:Computed tomography pulmonary angiogram (CTPA) is one of the most commonly ordered and frequently overused tests. The purpose of this study was to evaluate the mean radiation dose to patients getting CTPA and to identify factors that are associated with higher dose.Material and Methods:This institutionally approved retrospective study included all patients who had a CTPA to rule out acute pulmonary embolism between 2016 and 2018 in a tertiary care center. Patient data (age, sex, body mass index [BMI], and patient location), CT scanner type, image reconstruction methodology, and radiation dose parameters (dose-length product [DLP]) were recorded. Effective dose estimates were obtained by multiplying DLP by conversion coefficient (0.014 mSv•mGy−1•cm−1). Multivariate logistic regression analysis was performed to determine the factors affecting the radiation dose.Results:There were 2342 patients (1099 men and 1243 women) with a mean age of 58.1 years (range 0.2–104.4 years) and BMI of 31.3 kg/m2 (range 12–91.5 kg/m2). The mean effective radiation dose was 5.512 mSv (median – 4.27 mSv; range 0.1–43.0 mSv). Patient factors, including BMI >25 kg/m2, male sex, age >18 years, and intensive care unit (ICU) location, were associated with significantly higher dose (P < 0.05). CT scanning using third generation dual-source scanner with model-based iterative reconstruction (IR) had significantly lower dose (mean: 4.90 mSv) versus single-source (64-slice) scanner with filtered back projection (mean: 9.29 mSv, P < 0.001).Conclusion:Patients with high BMI and ICU referrals are associated with high CT radiation dose. They are most likely to benefit by scanning on newer generation scanner using advance model-based IR techniques.

Image reconstruction based on compressed sensing for sparse-data endoscopic photoacoustic tomography

Endoscopic photoacoustic tomography (EPAT) is an interventional application of photoacoustic tomography (PAT) to visualize anatomical features and functional components of biological cavity structures such as nasal cavity, digestive tract or coronary arterial vessels. One of the main challenges in clinical applicability of EPAT is the incomplete acoustic measurements due to the limited detectors or the limited-view acoustic detection enclosed in the cavity. In this case, conventional image reconstruction methodologies suffer from significantly degraded image quality. This work introduces a compressed-sensing (CS)-based method to reconstruct a high-quality image that represents the initial pressure distribution on a luminal cross-section from incomplete discrete acoustic measurements. The method constructs and trains a complete dictionary for the sparse representation of the photoacoustically-induced acoustic measurements. The sparse representation of the complete acoustic signals is then optimally obtained based on the sparse measurements and a sensing matrix. The complete acoustic signals are recovered from the sparse representation by inverse sparse transformation. The image of the initial pressure distribution is finally reconstructed from the recovered complete signals by using the time reversal (TR) algorithm. It was shown with numerical experiments that high-quality images with reduced under-sampling artifacts can be reconstructed from sparse measurements. The comparison results suggest that the proposed method outperforms the standard TR reconstruction by 40% in terms of the structural similarity of the reconstructed images.

AN EXPLORATION TO IMAGE RECONSTRUCTION APPROACHES TO REPAIR REAL TIME IMAGES

When the image is captured from the source using some scanner or the camera, it can have number of deficiencies because of the camera, environment or lack of expertise problems. These problems results distorted capturing of images. But to perform any image processing operation, it is required to resolve these impurities over the image. In this paper, the importance of image restoration is defined against various problems of images. The paper has discussed some of effective image reconstruction methodologies.

The influence of nanoparticles on dendritic grain growth in Mg alloys

Melt processing offers a cost effective method for producing metal matrix nanocomposite (MMNC) components; however, the influence of nanoparticles on the evolving microstructure during solidification is still not well understood. In this study, the effect of SiC nanoparticles on α-Mg dendrite evolution in a Mg-25Zn-7Al (wt.%) alloy was investigated through 4D (three dimensions plus time) synchrotron tomographic quantification of solidification experiments conducted at different cooling rates with and without nanoparticles. Key features of the solidifying primary α-Mg dendritic grains were quantified, including grain morphology, size distribution, and dendrite tip velocity. To obtain the high-contrast tomography dataset necessary for structure quantification, a new image reconstruction and processing methodology was implemented. The results reveal that the addition of nanoparticles increases grain nucleation whilst restricting dendritic growth and altering the dendritic grain growth morphology. Using LGK model calculations, it is shown that these changes in solidification microstructure occur as a result of nanoparticle-induced restriction in Zn's effective diffusivity ahead of the dendrite tips, reducing tip velocity. The results both suggest the key phenomena required to be simulated when numerically modelling solidifying Mg-based MMNC and provide the data required to validate those models.

Abstracts of the 33rd International Austrian Winter Symposium

K. Binzel1, A. Adelaja1, C. L. Wright1, D. Scharre2, J. Zhang1, M. V. Knopp11Wright Center of Innovation in Biomedical Imaging, Department of Radiology, The Ohio State University Wexner Medical Center, Columbus, OH, USA; 2Department of Neurology, The Ohio State University Wexner Medical Center, Columbus, OH, USACorrespondence: K. Binzel Aim: Digital photon counting allows for higher definition image reconstruction than on conventional PET systems due to increased system sensitivity, time of flight timing resolution, and count densities (1). For neurologic applications, these improvements may greatly increase the utility and accuracy of PET imaging. We validated and optimized high definition image reconstruction methodologies using the Hoffman brain phantom. Methods: A Hoffman brain phantom was filled with 52 MBq 18F-FDG and imaged on a pre-commercial release digital photon counting PET/CT (Philips Vereos). Five minute acquisitions were repeated over several runs using the dedicated brain 256 mm field of view (FOV) and the whole body 576 mm FOV. Listmode data from each acquisition was reconstructed using the high definition (HD) 2x2x2 mm voxel matrix. Reconstructions were performed with 3 iterations and a range of subsets, 21, 17, 13, and 9. Additionally, the use of system point spread function (PSF) correction and/or a Gaussian filter were enabled. Regions of interest (ROIs) were placed in 10 unique regions of the phantom for quantitative assessment. Blinded reader reviews were performed to assess image quality. Results: We found that the optimal reconstruction settings for each FOV were distinct. For the 576 mm FOV the use of PSF alone, no Gaussian filter, gave the most accurate quantitative results. The addition of the Gaussian filter resulted in underestimation of activity concentrations. The average recovery coefficients (RCs) of all ROIs were very similar among reconstructions with different numbers of subsets. The average RCs were 0.97, 0.95, 0.94, and 0.92 for the 21, 17, 13, and 9 subsets reconstructions, respectively. Blinded review conveyed that the 13 subset images were most preferable with regard to contrast and image noise. Thus the HD reconstruction with PSF only using 3 iterations and 13 subsets was optimal for the whole body FOV acquisitions. For the 256 mm FOV a Gaussian filter was used in reconstruction as the PSF alone lead to overestimation of activity concentrations. As with the whole body FOV, each subset setting gave similar quantitative results, average RCs were 1.02, 1.01, 0.98, and 0.95 for the 21, 17, 13, and 9 subset reconstructions. Image review again showed that fewer subsets were preferred, with the 9 subset reconstruction now being most ideal. For acquisitions with the dedicated brain FOV, HD reconstruction with both PSF correction and a Gaussian filter using 3 iterations with 9 subsets is optimal for imaging with FDG. Conclusion: Optimization of Neuro-PET reconstruction settings revealed that these setting must be tailored to acquisition characteristics, namely the chosen field of view. Phantom validation demonstrated that high quantitative accuracy and excellent image quality is readily achieved for neurologic imaging on next-generation digital PET systems when using either a dedicated brain field of view or the wider whole body field of view. Reference 1. Wright CL, Binzel K, Zhang J, Knopp MV. Advanced Functional Tumor Imaging and Precision Nuclear Medicine Enabled by Digital PET Technologies. Contrast Media & Molecular Imaging. 2017;2017:5260305. https://doi.org/10.1155/2017/5260305.

Evaluation of high-quality image reconstruction techniques applied to high-resolution Z-contrast imaging

High-quality image reconstruction techniques allow the generation of high pixel density images from a set of low-resolution micrographs. In general, these techniques consist of two main steps, namely, accurate registration, and formulation of an appropriate forward image model via some restoration method. There exist a wide variety of algorithms to cope with both stages and depending on their practical applications, some methods can outperform others, since they can be sensitive to the assumed data model, noise, drift, etc. When dealing with images generated by Z-contrast scanning transmission electron microscopes, a current trend is based on non-rigid approximations in the registration stage. In our work we aimed at reaching similar accuracy but addressing the most complex calculations in the reconstruction stage, instead of in the registration stage (as the non-rigid approaches do), but using a much smaller number of images. We review some of the most significant methods and address their shortcomings when they are applied to the field of microscopy. Simulated images with known targets will be used to evaluate and compare the main approaches in terms of quality enhancement and computing time. In addition, a procedure to determine the reference image will be proposed to minimise the global drift on the series. The best registration and restoration strategies will be applied to experimental images in order to point up the enhanced capability of this high quality image reconstruction methodology in this field.

Three Dimensional MR Brain Image Reconstruction and Interactive Segmentation Based Tumor Region Extraction Using Active Contour Models

&lt;p&gt;Magnetic resonance imaging (MRI) system projects the anatomical structures of brain onto&lt;br /&gt;spatial domain. Three-dimensional (3D) construction of brain tumor using several slides of MRI has&lt;br /&gt;always been of keen interest for diagnosis and for research purpose. This paper presents a three&lt;br /&gt;dimensional brain image reconstruction methodology which is very much helpful in predicting the depth&lt;br /&gt;of any abnormality in the image and also proposes an active contour based approach for the extraction of&lt;br /&gt;the abnormal (tumor) region. The extracted regions are modeled using Iso surface 3D view models.&lt;/p&gt;

Performance evaluation of 18 F radioluminescence microscopy using computational simulation.

Radioluminescence microscopy can visualize the distribution of beta-emitting radiotracers in live single cells with high resolution. Here, we perform a computational simulation of 18 F positron imaging using this modality to better understand how radioluminescence signals are formed and to assist in optimizing the experimental setup and image processing. First, the transport of charged particles through the cell and scintillator and the resulting scintillation is modeled using the GEANT4 Monte-Carlo simulation. Then, the propagation of the scintillation light through the microscope is modeled by a convolution with a depth-dependent point-spread function, which models the microscope response. Finally, the physical measurement of the scintillation light using an electron-multiplying charge-coupled device (EMCCD) camera is modeled using a stochastic numerical photosensor model, which accounts for various sources of noise. The simulated output of the EMCCD camera is further processed using our ORBIT image reconstruction methodology to evaluate the endpoint images. The EMCCD camera model was validated against experimentally acquired images and the simulated noise, as measured by the standard deviation of a blank image, was found to be accurate within 2% of the actual detection. Furthermore, point source simulations found that a reconstructed spatial resolution of 18.5 μm can be achieved near the scintillator. As the source is moved away from the scintillator, spatial resolution degrades at a rate of 3.5 μm per μm distance. These results agree well with the experimentally measured spatial resolution of 30-40 μm (live cells). The simulation also shows that the system sensitivity is 26.5%, which is also consistent with our previous experiments. Finally, an image of a simulated sparse set of single cells is visually similar to the measured cell image. Our simulation methodology agrees with experimental measurements taken with radioluminescence microscopy. This in silico approach can be used to guide further instrumentation developments and to provide a framework for improving image reconstruction.

Impact and correction of the bladder uptake on 18F-FCH PET quantification: a simulation study using the XCAT2 phantom

The spill-in counts from neighbouring regions can significantly bias the quantification over small regions close to high activity extended sources. This effect can be a drawback for 18F-based radiotracers positron emission tomography (PET) when quantitatively evaluating the bladder area for diseases such as prostate cancer. In this work, we use Monte Carlo simulations to investigate the impact of the spill-in counts from the bladder on the quantitative evaluation of prostate cancer when using 18F-Fluorcholine (FCH) PET and we propose a novel reconstruction-based correction method. Monte Carlo simulations of a modified version of the XCAT2 anthropomorphic phantom with 18F-FCH biological distribution, variable bladder uptake and inserted prostatic tumours were used in order to obtain simulated realistic 18F-FCH data. We evaluated possible variations of the measured tumour Standardized Uptake Value (SUV) for different values of bladder uptake and propose a novel correction by appropriately adapting image reconstruction methodology. The correction is based on the introduction of physiological background terms on the reconstruction, removing the contribution of the bladder to the final image. The bladder is segmented from the reconstructed image and then forward-projected to the sinogram space. The resulting sinograms are used as background terms for the reconstruction. SUVmax and SUVmean could be overestimated by 41% and 22% respectively due to the accumulation of radiotracer in the bladder, with strong dependence on bladder-to-lesion ratio. While the SUVs measured under these conditions are not reliable, images corrected using the proposed methodology provide better repeatability of SUVs, with biases below 6%. Results also showed remarkable improvements on visual detectability. The spill-in counts from the bladder can affect prostatic SUV measurements of 18F-FCH images, which can be corrected to less than 6% using the proposed methodology, providing reliable SUV values even in the presence of high radioactivity accumulation in the bladder.

Image Processing in Radiation Therapy

This article reviews Image Processing in Radiation Therapy by Kristy Brock Editor: Series Editor: William R. Hendee , Boca Raton, FL, 2013. Hardcover 286 pp. Price: $103.96. ISBN: 9781439830178.

A novel neural network based image reconstruction model with scale and rotation invariance for target identification and classification for Active millimetre wave imaging

Millimetre wave imaging (MMW) is gaining tremendous interest among researchers, which has potential applications for security check, standoff personal screening, automotive collision-avoidance, and lot more. Current state-of-art imaging techniques viz. microwave and X-ray imaging suffers from lower resolution and harmful ionizing radiation, respectively. In contrast, MMW imaging operates at lower power and is non-ionizing, hence, medically safe. Despite these favourable attributes, MMW imaging encounters various challenges as; still it is very less explored area and lacks suitable imaging methodology for extracting complete target information. Keeping in view of these challenges, a MMW active imaging radar system at 60 GHz was designed for standoff imaging application. A C-scan (horizontal and vertical scanning) methodology was developed that provides cross-range resolution of 8.59 mm. The paper further details a suitable target identification and classification methodology. For identification of regular shape targets: mean-standard deviation based segmentation technique was formulated and further validated using a different target shape. For classification: probability density function based target material discrimination methodology was proposed and further validated on different dataset. Lastly, a novel artificial neural network based scale and rotation invariant, image reconstruction methodology has been proposed to counter the distortions in the image caused due to noise, rotation or scale variations. The designed neural network once trained with sample images, automatically takes care of these deformations and successfully reconstructs the corrected image for the test targets. Techniques developed in this paper are tested and validated using four different regular shapes viz. rectangle, square, triangle and circle.

Taylor series expansion based multidimensional image reconstruction for confocal and 4pi microscopy

We propose and experimentally demonstrate a three-dimensional (3D) image reconstruction methodology based on Taylor series approximation (TSA) in a Bayesian image reconstruction formulation. TSA incorporates the requirement of analyticity in the image domain, and acts as a finite impulse response filter. This technique is validated on images obtained from widefield, confocal laser scanning fluorescence microscopy and two-photon excited 4pi (2PE-4pi) fluorescence microscopy. Studies on simulated 3D objects, mitochondria-tagged yeast cells (labeled with Mitotracker Orange) and mitochondrial networks (tagged with Green fluorescent protein) show a signal-to-background improvement of 40% and resolution enhancement from 360 to 240 nm. This technique can easily be extended to other imaging modalities (single plane illumination microscopy (SPIM), individual molecule localization SPIM, stimulated emission depletion microscopy and its variants).

Sparse-Coding-Based Computed Tomography Image Reconstruction

Computed tomography (CT) is a popular type of medical imaging that generates images of the internal structure of an object based on projection scans of the object from several angles. There are numerous methods to reconstruct the original shape of the target object from scans, but they are still dependent on the number of angles and iterations. To overcome the drawbacks of iterative reconstruction approaches like the algebraic reconstruction technique (ART), while the recovery is slightly impacted from a random noise (small amount of ℓ 2 norm error) and projection scans (small amount of ℓ 1 norm error) as well, we propose a medical image reconstruction methodology using the properties of sparse coding. It is a very powerful matrix factorization method which each pixel point is represented as a linear combination of a small number of basis vectors.

Aberration correction for transcranial photoacoustic tomography of primates employing adjunct image data.

A challenge in photoacoustic tomography (PAT) brain imaging is to compensate for aberrations in the measured photoacoustic data due to their propagation through the skull. By use of information regarding the skull morphology and composition obtained from adjunct x-ray computed tomography image data, we developed a subject-specific imaging model that accounts for such aberrations. A time-reversal-based reconstruction algorithm was employed with this model for image reconstruction. The image reconstruction methodology was evaluated in experimental studies involving phantoms and monkey heads. The results establish that our reconstruction methodology can effectively compensate for skull-induced acoustic aberrations and improve image fidelity in transcranial PAT.

Fast-updating and nonrepeating Lissajous image reconstruction method for capturing increased dynamic information

We present a fast-updating Lissajous image reconstruction methodology that uses an increased image frame rate beyond the pattern repeat rate generally used in conventional Lissajous image reconstruction methods. The fast display rate provides increased dynamic information and reduced motion blur, as compared to conventional Lissajous reconstruction, at the cost of single-frame pixel density. Importantly, this method does not discard any information from the conventional Lissajous image reconstruction, and frames from the complete Lissajous pattern can be displayed simultaneously. We present the theoretical background for this image reconstruction methodology along with images and video taken using the algorithm in a custom-built miniaturized multiphoton microscopy system.

Task-oriented quantitative image reconstruction in emission tomography for single- and multi-subject studies

Task-based selection of image reconstruction methodology in emission tomography is a critically important step when designing a PET study. This paper concerns optimizing, given the measured data of the study only, reconstruction performance for a range of quantification tasks: finding the mean radioactivity concentration for different regions of interests (ROIs), different ROI sizes and different group sizes (i.e. the number of subjects in the PET study). At present, the variability of quantification performance of different reconstruction methods, according to both the ROI and group sizes, is largely ignored. In this paper, it is shown that both the ROI and group size have a tremendous impact on the error of the estimator for the task of ROI quantification. A study-specific, task-oriented and space-variant selection rule is proposed that selects a close to optimal estimate drawn from a series of estimates obtained by filtered backprojection (FBP) and different OSEM (ordered subset expectation maximization) iterations. The optimality criterion is to minimize an estimated mean square error (MSE), where the MSE is estimated from the data in the study using the bootstrap resampling technique. The proposed approach is appropriate for both pixel-level estimates and ROI estimates in single- and multi-subject studies. An extensive multi-trial simulation study using a 2D numerical phantom and relevant count levels shows that the proposed selection rule can produce quantitative estimates that are close to the estimates that minimize the true MSE (where the true MSE can only be obtained from many independent Monte-Carlo realizations with knowledge of the ground truth). This indicates that with the proposed selection rule one can obtain a close to optimal estimate while avoiding the critical step of selecting user-defined reconstruction settings (such as an OSEM iteration number or the choice between FBP and OSEM). In this initial 2D study, only FBP and OSEM reconstruction methods are considered but the proposed selection rule should readily generalize to different estimators (i.e. different reconstruction algorithms) and 3D imaging.

Phase-sensitive black-blood coronary vessel wall imaging

Black-blood MR coronary vessel wall imaging may become a powerful tool for the quantitative and noninvasive assessment of atherosclerosis and positive arterial remodeling. Although dual-inversion recovery is currently the gold standard, optimal lumen-to-vessel wall contrast is sometimes difficult to obtain, and the time window available for imaging is limited due to competing requirements between blood signal nulling time and period of minimal myocardial motion. Further, atherosclerosis is a spatially heterogeneous disease, and imaging at multiple anatomic levels of the coronary circulation is mandatory. However, this requirement of enhanced volumetric coverage comes at the expense of scanning time. Phase-sensitive inversion recovery has shown to be very valuable for enhancing tissue-tissue contrast and for making inversion recovery imaging less sensitive to tissue signal nulling time. This work enables multislice black-blood coronary vessel wall imaging in a single breath hold by extending phase-sensitive inversion recovery to phase-sensitive dual-inversion recovery, by combining it with spiral imaging and yet relaxing constraints related to blood signal nulling time and period of minimal myocardial motion.