Tomographic Image Reconstruction Research Articles

A Joint-Parameter Estimation and Bayesian Reconstruction Approach to Low-Dose CT.

Most penalized maximum likelihood methods for tomographic image reconstruction based on Bayes' law include a freely adjustable hyperparameter to balance the data fidelity term and the prior/penalty term for a specific noise-resolution tradeoff. The hyperparameter is determined empirically via a trial-and-error fashion in many applications, which then selects the optimal result from multiple iterative reconstructions. These penalized methods are not only time-consuming by their iterative nature, but also require manual adjustment. This study aims to investigate a theory-based strategy for Bayesian image reconstruction without a freely adjustable hyperparameter, to substantially save time and computational resources. The Bayesian image reconstruction problem is formulated by two probability density functions (PDFs), one for the data fidelity term and the other for the prior term. When formulating these PDFs, we introduce two parameters. While these two parameters ensure the PDFs completely describe the data and prior terms, they cannot be determined by the acquired data; thus, they are called complete but unobservable parameters. Estimating these two parameters becomes possible under the conditional expectation and maximization for the image reconstruction, given the acquired data and the PDFs. This leads to an iterative algorithm, which jointly estimates the two parameters and computes the to-be reconstructed image by maximizing a posteriori probability, denoted as joint-parameter-Bayes. In addition to the theoretical formulation, comprehensive simulation experiments are performed to analyze the stopping criterion of the iterative joint-parameter-Bayes method. Finally, given the data, an optimal reconstruction is obtained without any freely adjustable hyperparameter by satisfying the PDF condition for both the data likelihood and the prior probability, and by satisfying the stopping criterion. Moreover, the stability of joint-parameter-Bayes is investigated through factors such as initialization, the PDF specification, and renormalization in an iterative manner. Both phantom simulation and clinical patient data results show that joint-parameter-Bayes can provide comparable reconstructed image quality compared to the conventional methods, but with much less reconstruction time. To see the response of the algorithm to different types of noise, three common noise models are introduced to the simulation data, including white Gaussian noise to post-log sinogram data, Poisson-like signal-dependent noise to post-log sinogram data and Poisson noise to the pre-log transmission data. The experimental outcomes of the white Gaussian noise reveal that the two parameters estimated by the joint-parameter-Bayes method agree well with simulations. It is observed that the parameter introduced to satisfy the prior's PDF is more sensitive to stopping the iteration process for all three noise models. A stability investigation showed that the initial image by filtered back projection is very robust. Clinical patient data demonstrated the effectiveness of the proposed joint-parameter-Bayes and stopping criterion.

A novel figure of merit to investigate 68Ga PET/CT image quality based on patient weight and lesion size using Q.Clear reconstruction algorithm: A phantom study

IntroductionQ.Clear is a Bayesian penalised-likelihood algorithm that uses a β-value for positron emission tomography(PET)/computed tomography(CT) image reconstruction(IR). Our study proposes a novel figure of merit, named CRBV, to compare the Q.Clear performances using 68Ga PET/CT image with the ordered-subset-expectation–maximization(OSEM) algorithm and to identify the optimal β-values for these images using two phantoms mimicking normal and overweight patients. MethodsNEMA IQ phantom with or without a ring of water-filled plastic bags (NEMAstd and NEMAow, respectively) was acquired and reconstructed with OSEM and Q.Clear at various β-values and minutes/bed position(min/bp). Contrast recovery(CR), background variability(BV) and CRBV were calculated. Highest CRBV values were used to identify optimal β-value ranges. ResultsQ.Clear with 250 ≤ β ≤ 800 improved CRBV compared to OSEM for all the investigated spheres and acquisition setups. Outside of this range, Q.Clear still outperformed OSEM with few exceptions depending on spheres diameters and phantoms(e.g.,β-value = 1600 for diameters ≤ 17 mm using the NEMAow phantom). Regarding the CRBV performance for IR optimization, for the 4 min/bp NEMAstd IR, β-values = 300 ÷ 350 allowed to simultaneously optimize all diameters(except for the 10 mm); for the NEMAow IR, β-values = 350 ÷ 500 were needed for diameters > 20 mm, while β-values = 200 ÷ 250 were selected for the remaining diameters. For the 2 min/bp, β-value = 500 was suitable for diameters > 17 mm in both NEMAstd and NEMAow IR, while for smaller diameters β-value = 200 and β-values = 250 ÷ 350 were obtained for NEMAstd and NEMAow, respectively. ConclusionAlmost all tested β-values of Q.Clear improved the CRBV compared to OSEM. In both phantoms, simulating normal and over-weight patients, optimal β-values were found according to lesion sizes and investigated acquisition times.

A High Accuracy Voltage Approximation Model Based on Object-oriented Sensitivity Matrix Estimation (OO-SME Model) in Electrical Impedance Tomography.

The image reconstruction in electrical impedance tomography (EIT) has low accuracy due to the approximation error between the measured voltage change and the approximated voltage change, from which the object cannot be accurately reconstructed and quantitatively evaluated. A voltage approximation model based on object-oriented sensitivity matrix estimation (OO-SME model) is proposed to reconstruct the image with high accuracy. In the OO-SME model, a sensitivity matrix of the object-field is estimated, and the sensitivity matrix change from the background-field to the object-field is estimated to optimize the approximated voltage change, from which the approximation error is eliminated to improve the reconstruction accuracy. Against the existing linear and nonlinear models, the approximation error in the OO-SME model is eliminated, thus an image with higher accuracy is reconstructed. The simulation shows that the OO-SME model reconstructs a more accurate image than the existing models for quantitative evaluation. The relative accuracy (RA) of reconstructed conductivity is increased up to 83.98% on average. The experiment of lean meat mass evaluation shows that the RA of lean meat mass is increased from 7.70% with the linear model to 54.60% with the OO-SME model. It is concluded that the OO-SME model reconstructs a more accurate image to evaluate the object quantitatively than the existing models.

Combining band-frequency separation and deep neural networks for optoacoustic imaging

In this paper we consider the problem of image reconstruction in optoacoustic tomography. In particular, we devise a deep neural architecture that can explicitly take into account the band-frequency information contained in the sinogram. This is accomplished by two means. First, we jointly use a linear filtered back-projection method and a fully dense UNet for the generation of the images corresponding to each one of the frequency bands considered in the separation. Secondly, in order to train the model, we introduce a special loss function consisting of three terms: (i) a separating frequency bands term; (ii) a sinogram-based consistency term and (iii) a term that directly measures the quality of image reconstruction and which takes advantage of the presence of ground-truth images present in training dataset. Numerical experiments show that the proposed model, which can be easily trainable by standard optimization methods, presents an excellent generalization performance quantified by a number of metrics commonly used in practice. Also, in the testing phase, our solution has a comparable (in some cases lower) computational complexity, which is a desirable feature for real-time implementation of optoacoustic imaging.

GPU-accelerated iterative method for FD-OCT image reconstruction with an image-level cross-domain regularizer.

The image reconstruction for Fourier-domain optical coherence tomography (FD-OCT) could be achieved by iterative methods, which offer a more accurate estimation than the traditional inverse discrete Fourier transform (IDFT) reconstruction. However, the existing iterative methods are mostly A-line-based and are developed on CPU, which causes slow reconstruction. Besides, A-line-based reconstruction makes the iterative methods incompatible with most existing image-level image processing techniques. In this paper, we proposed an iterative method that enables B-scan-based OCT image reconstruction, which has three major advantages: (1) Large-scale parallelism of the OCT dataset is achieved by using GPU acceleration. (2) A novel image-level cross-domain regularizer was developed, such that the image processing could be performed simultaneously during the image reconstruction; an enhanced image could be directly generated from the OCT interferogram. (3) The scalability of the proposed method was demonstrated for 3D OCT image reconstruction. Compared with the state-of-the-art (SOTA) iterative approaches, the proposed method achieves higher image quality with reduced computational time by orders of magnitude. To further show the image enhancement ability, a comparison was conducted between the proposed method and the conventional workflow, in which an IDFT reconstructed OCT image is later processed by a total variation-regularized denoising algorithm. The proposed method can achieve a better performance evaluated by metrics such as signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR), while the speed is improved by more than 30 times. Real-time image reconstruction at more than 20 B-scans per second was realized with a frame size of 4096 (axial) × 1000 (lateral), which showcases the great potential of the proposed method in real-world applications.

Comparative study of sex estimates in adult skulls using direct measurement and tomographic image reconstruction.

Sex estimation is an important procedure in forensic anthropology for human identification. The development of new technologies, such as three-dimensional computed tomography (CT), has provided excellent alternatives for this purpose. This study examined and compared a morphological method for sex estimation using two different approaches - direct measurement of physical structures and tomographic analysis using 3D images. A total of 111 skulls from the Museum of Human Anatomy at the University of São Paulo (MAH-USP) were used, (60 males and 51 females). All specimens were scanned by Philips Brilliance 64 CT scanner equipment and their corresponding images were reconstructed in three-dimensional (3D) models. The morphological characteristics of the skulls were analyzed by an observer who was blinded to the sex of the specimens. Five cranial structures were analyzed: external occipital crest, mastoid process, supraorbital margin, glabella, and mental eminence. The structures were scored 1 to 5 according to Buikstra and Ubelaker and validated by Walker. The success rates of the sex estimates obtained through direct measurement of the dry skulls ranged from 67.4% to 70.4% as compared to 60.2% to 68.1% for CT reconstruction. When analyzed separately, the maximum accuracy of the method was 68.33% in males and 88.24% in females in the physical analysis of structures. The glabella and mastoid process were the most effective structures to estimate sex through both techniques, respectively. Our results show that 3D CT images can be accurately used in the morphological analysis for sex estimation, representing a viable alternative in forensic anthropology.

Deep Embedding-Attention-Refinement for Sparse-View CT Reconstruction

Tomographic image reconstruction with deep learning is an emerging field of applied artificial intelligence. Reducing radiation dose with sparse views’ reconstruction is a significant task in cardiac imaging. Many efforts are contributing to sparse-view tomography imaging, but it is still a challenge for achieving good images from high sparse-view level, such as 60 views. In this study, we proposed a Deep Embedding-Attention-Refinement (DEAR) network to fundamentally address this challenge. DEAR consists of three modules including deep embedding, deep attention, and deep refinement. The measurement is extended by deep embedding network to generate artifact-reduction images. Then, the deep attention network is employed to remove sparse-view artifacts and correct wrong details introduced by deep embedding network. Finally, the deep refinement module is used to refine finer image features and structures. The results on clinical datasets demonstrate the efficiency of our proposed DEAR in edge preservation and feature recovery.

Comparison of iteratively regularized Gauss-Newton method with Adam optimization for image reconstruction in electrical impedance tomography

Comparison of iteratively regularized Gauss-Newton method with Adam optimization for image reconstruction in electrical impedance tomography

Dilated POCS: Minimax Convex Optimization

<italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Alternating projection onto convex sets</i> (POCS) provides an iterative procedure to find a signal that satisfies two or more convex constraints when the sets intersect. For nonintersecting constraints, the method of simultaneous projections produces a minimum mean square error (MMSE) solution. In certain cases, a minimax solution is more desirable. Generating a minimax solution is possible using dilated POCS. The minimax solution uses morphological dilation of nonintersecting signal convex constraints. The sets are progressively dilated to the point where there is intersection at a minimax solution. Examples are given contrasting the MMSE and minimax solutions in problems of tomographic reconstruction of images. Dilated POCS adds a new imaging modality for image synthesis. Lastly, morphological erosion of signal sets is suggested as a method to shrink the overlap when sets intersect at more than one point.

A scalable neural network architecture for self-supervised tomographic image reconstruction

We present a lightweight and scalable artificial neural network architecture which is used to reconstruct a tomographic image from a given sinogram.

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.

Tomographic image reconstruction techniques for accurate spent fuel assembly verification

Non-proliferation and the security of nuclear materials are essential. The international atomic energy agency (IAEA) considers a tomographic image acquisition technique of spent fuel assemblies a promising technique to accurately verify the rod-by-rod spent fuel conditions stored in a water pool. Researchers at Yonsei University in Korea developed the bismuth germanate (BGO) scintillator-based Yonsei Single-photon Emission Computed Tomography (YSECT). Previous research validated the YSECT system experimentally to quickly evaluate the radioactivity distribution of test fuel rods in the Korea Institute of Nuclear Nonproliferation and Control (KINAC). Quick verification of the fuel assembly requires the development of a high-quality image reconstruction algorithm that enables image acquisition within a short time. This study examined various tomographic image reconstruction techniques to identify patterns of missing fuel rods accurately. Rotational projection image data sets were obtained for 15 patterns of test fuel rods for 900 seconds using the single-photon emission computed tomography (SPECT) system installed at KINAC. The projection images were acquired every 5° while four 64-channel detectors rotated 90°. The acquired images were reconstructed using the following methods: filtered back-projection, simultaneous iterative reconstruction technique, order-subset simultaneous algebraic reconstruction technique, maximum likelihood expectation maximization (MLEM), and Fast-Iterative Shrinkage-Thresholding algorithm (FISTA). Among the reconstruction algorithms used in this study, the image quality of MLEM showed the best performance, and the image contrast of FISTA showed the highest result. Therefore, the signal-to-noise ratio of the tomographic image was improved using the image reconstruction technique optimized for the YSECT system to verify the patterns of fuel rods. Hence, even for the low-quality measured data with the short-time scan of the SPECT system, this advanced technique is expected to show better discriminability of the patterns of fuel rods in the assembly.

Thoracic cavity remodeling and pulmonary function change after chest wall resection.

Chest wall resection (CWR) is an essential procedure for treating malignancies and infectious conditions of the chest wall. However, there are few studies on the pulmonary function and changes in thoracic cavity volume (TCV) related to CWR. This study aims to investigate the effects of CWR on long-term changes in TCV and pulmonary function. Data of patients who underwent CWR between 2001 and 2021 were retrospectively reviewed. Patients who underwent single rib or lung resection rather than wedge resection were excluded. TCV (liter) was defined as the sum of the right and left TCVs (RCV and LCV) and was measured using computed tomography image reconstruction software. Changes in pulmonary function and TCV 1 year postoperatively were analyzed. A total of 45 patients were included. The number of resected ribs was 2 in 16 (35.6%) and ≥3 in 29 (64.4%) patients. Thirty patients underwent reconstruction. Long-term post-CWR decreased in forced vital capacity (FVC) (-7.9%, P=0.004) and forced expiratory volume in 1 second (FEV1) (-7.0%, P=0.002) were significant. There was no significant decrease in FEV1/FVC ratio (-3.0%, P=0.06), diffusing capacity of the lung for carbon monoxide (DLCO) (-5.9%, P=0.18) and TCV (-3.1%, P=0.10). There was no correlation between changes in TCV and decreases in FVC (r=0.12, P=0.56) or FEV1 (r=0.15, P=0.45). After right-side CWR (n=27), RCV (-7.8%, P=0.01) decreased significantly, whereas LCV (+2.1%, P=0.58) did not. The left-side CWR exhibited an identical pattern. (LCV: -8.5%, P=0.004; RCV: +1.3%, P=0.85). In the ≥3 rib-resection group, FVC (-9.5%, P=0.02), FEV1 (-7.9%, P=0.02) and TCV (-6.4%, P=0.04) decreased significantly. No significant changes were noted in the 2 rib-resection group. There were no significant differences in the changes in pulmonary function nor TCV between the reconstruction and no-reconstruction groups. The long-term decrease in pulmonary function after CWR was significant, especially after ≥3-rib resection.

Uncertainty Quantification in Intelligent-Based Electrical Resistivity Tomography Image Reconstruction With Monte Carlo Dropout Strategy

The Electrical Resistivity Tomography (ERT) method plays an essential role in researching electrical properties at the core scale, and the resistivity image inversion reconstruction technique is the key to constraining measurement accuracy. In recent years, with the improvement of computer operation capabilities and the acquisition of a large amount of geophysical data, the inversion algorithms represented by machine learning (ML) have made it possible to reconstruct ERT images automatically. However, the solution of ERT image inversion has large uncertainty due to high nonlinearity. Meanwhile, traditional ML methods are not initially developed for evaluating the uncertainty of such reconstruction results. In this paper, we propose a novel MC-Net ML scheme to quantitatively estimate the uncertainty of the ERT image reconstruction by introducing the Monte-Carlo Dropout strategy with a multiple stochastic method to approximate Bayesian inference. As one important evaluation index, the correlation coefficient (CC) between target image and reconstructed image is used to compare the effects of six types of simulated data carried by different ML schemes. The inversion results show that MC-Net increases lower limit of CC on the premise of ensuring average CC. Moreover, the accuracy of regression and classification results defined in this paper also obtained the highest value by MC-Net. With the proposed scheme, we can further provide a quantification of the uncertainty to determine whether reconstructed images are reliable after reconstructing the ERT images, which is of great significance for actual intelligent inversion tasks.

Electrical Impedance Tomography Image Reconstruction With Attention-Based Deep Convolutional Neural Network

Electrical impedance tomography (EIT) is a promising functional and structural imaging method in process tomography. However, due to the ’soft-field’ nature and the high dependence on the prior information, it often suffers serious artifacts in quantitative analysis. Most recently, EIT image quality has improved significantly because of the state-of-the-art deep learning-based models in the aspect of solving the inverse problem, especially fully convolutional networks (FCN) and V-Net variants. Despite their success, these deep convolutional networks (CNNs) have two limitations: (1) The long-range information transition is frequently lost and the reverse gradient often disappears in deep CNNs; (2) Some novel skip connections, such as residual and dense connections, often occupy substantial computational resources. To overcome these two limitations, we propose V <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> A-Net, a new neural architecture based on redesigned feature transited connections by the terms of (1) A pre-reconstructor based on the iterative Newton-Raphson method, which maps the nonlinear function between the measurements and the initial images, (2) Dual cascaded V-Net are combined, which play the role of an encoder and a decoder, respectively, (3) A new parallel attention mechanism via channel attention and coordinate attention to learn the conductivity distributions and boundary-shaped feature separately, and (4) the light-weight skip connections reduce the computational resources (or accelerate the inference speed) of EIT imaging. The V <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> A-Net is evaluated by using the multi-phase flow industrial applications, and the results demonstrate that (1) V <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> A-Net has better performance in shape reconstruction with sharp ’corner’, (2) V <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> A-Net could reconstruct the model accurately where it has some low-contrast conductivity distributions, (3) V <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> A-Net enhances the quality of interfaces with the stratified flow, and (4) the pruned V <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> A-Net achieves significant speedup compared with the VDD-Net or V <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> DNet. The analyses show that the average relative error is 0.05, the average correlation coefficient is 0.92, the average structural similarity is 0.92 on the testing datasets. In addition, the average relative cover ratio is 0.97 and the average relative contrast ratio is 0.98 on the testing datasets.

Advances of deep learning in electrical impedance tomography image reconstruction.

Electrical impedance tomography (EIT) has been widely used in biomedical research because of its advantages of real-time imaging and nature of being non-invasive and radiation-free. Additionally, it can reconstruct the distribution or changes in electrical properties in the sensing area. Recently, with the significant advancements in the use of deep learning in intelligent medical imaging, EIT image reconstruction based on deep learning has received considerable attention. This study introduces the basic principles of EIT and summarizes the application progress of deep learning in EIT image reconstruction with regards to three aspects: a single network reconstruction, deep learning combined with traditional algorithm reconstruction, and multiple network hybrid reconstruction. In future, optimizing the datasets may be the main challenge in applying deep learning for EIT image reconstruction. Adopting a better network structure, focusing on the joint reconstruction of EIT and traditional algorithms, and using multimodal deep learning-based EIT may be the solution to existing problems. In general, deep learning offers a fresh approach for improving the performance of EIT image reconstruction and could be the foundation for building an intelligent integrated EIT diagnostic system in the future.

Phantom and clinical evaluation of the effect of a new Bayesian penalized likelihood reconstruction algorithm (HYPER Iterative) on 68Ga-DOTA-NOC PET/CT image quality

BackgroundBayesian penalized likelihood (BPL) algorithm is an effective way to suppress noise in the process of positron emission tomography (PET) image reconstruction by incorporating a smooth penalty. The strength of the smooth penalty is controlled by the penalization factor. The aim was to investigate the impact of different penalization factors and acquisition times in a new BPL algorithm, HYPER Iterative, on the quality of 68Ga-DOTA-NOC PET/CT images. A phantom and 25 patients with neuroendocrine neoplasms who underwent 68Ga-DOTA-NOC PET/CT were included. The PET data were acquired in a list-mode with a digital PET/CT scanner and reconstructed by ordered subset expectation maximization (OSEM) and the HYPER Iterative algorithm with seven penalization factors between 0.03 and 0.5 for acquisitions of 2 and 3 min per bed position (m/b), both including time-of-flight and point of spread function recovery. The contrast recovery (CR), background variability (BV) and radioactivity concentration ratio (RCR) of the phantom; The SUVmean and coefficient of variation (CV) of the liver; and the SUVmax of the lesions were measured. Image quality was rated by two radiologists using a five-point Likert scale.ResultsThe CR, BV, and RCR decreased with increasing penalization factors for four “hot” spheres, and the HYPER Iterative 2 m/b groups with penalization factors of 0.07 to 0.2 had equivalent CR and superior BV performance compared to the OSEM 3 m/b group. The liver SUVmean values were approximately equal in all reconstruction groups (range 5.95–5.97), and the liver CVs of the HYPER Iterative 2 m/b and 3 m/b groups with the penalization factors of 0.1 to 0.2 were equivalent to those of the OSEM 3 m/b group (p = 0.113–0.711 and p = 0.079–0.287, respectively), while the lesion SUVmax significantly increased by 19–22% and 25%, respectively (all p < 0.001). The highest qualitative score was attained at a penalization factor of 0.2 for the HYPER Iterative 2 m/b group (3.20 ± 0.52) and 3 m/b group (3.70 ± 0.36); those scores were comparable to or greater than that of the OSEM 3 m/b group (3.09 ± 0.36, p = 0.388 and p < 0.001, respectively).ConclusionsThe HYPER Iterative algorithm with a penalization factor of 0.2 resulted in higher lesion contrast and lower image noise than OSEM for 68Ga-DOTA-NOC PET/CT, allowing the same image quality to be achieved with less injected radioactivity and a shorter acquisition time.

Deep learning tomographic reconstruction through hierarchical decomposition of domain transforms.

Deep learning (DL) has shown unprecedented performance for many image analysis and image enhancement tasks. Yet, solving large-scale inverse problems like tomographic reconstruction remains challenging for DL. These problems involve non-local and space-variant integral transforms between the input and output domains, for which no efficient neural network models are readily available. A prior attempt to solve tomographic reconstruction problems with supervised learning relied on a brute-force fully connected network and only allowed reconstruction with a 1284 system matrix size. This cannot practically scale to realistic data sizes such as 5124 and 5126 for three-dimensional datasets. Here we present a novel framework to solve such problems with DL by casting the original problem as a continuum of intermediate representations between the input and output domains. The original problem is broken down into a sequence of simpler transformations that can be well mapped onto an efficient hierarchical network architecture, with exponentially fewer parameters than a fully connected network would need. We applied the approach to computed tomography (CT) image reconstruction for a 5124 system matrix size. This work introduces a new kind of data-driven DL solver for full-size CT reconstruction without relying on the structure of direct (analytical) or iterative (numerical) inversion techniques. This work presents a feasibility demonstration of full-scale learnt reconstruction, whereas more developments will be needed to demonstrate superiority relative to traditional reconstruction approaches. The proposed approach is also extendable to other imaging problems such as emission and magnetic resonance reconstruction. More broadly, hierarchical DL opens the door to a new class of solvers for general inverse problems, which could potentially lead to improved signal-to-noise ratio, spatial resolution and computational efficiency in various areas.

Impact of deep learning image reconstructions (DLIR) on coronary artery calcium quantification

BackgroundDeep learning image reconstructions (DLIR) have been recently introduced as an alternative to filtered back projection (FBP) and iterative reconstruction (IR) algorithms for computed tomography (CT) image reconstruction. The aim of this study was to evaluate the effect of DLIR on image quality and quantification of coronary artery calcium (CAC) in comparison to FBP.MethodsOne hundred patients were consecutively enrolled. Image quality–associated variables (noise, signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR)) as well as CAC-derived parameters (Agatston score, mass, and volume) were calculated from images reconstructed by using FBP and three different strengths of DLIR (low (DLIR_L), medium (DLIR_M), and high (DLIR_H)). Patients were stratified into 4 risk categories according to the Coronary Artery Calcium - Data and Reporting System (CAC-DRS) classification: 0 Agatston score (very low risk), 1–99 Agatston score (mildly increased risk), Agatston 100–299 (moderately increased risk), and ≥ 300 Agatston score (moderately-to-severely increased risk).ResultsIn comparison to standard FBP, increasing strength of DLIR was associated with a significant and progressive decrease of image noise (p < 0.001) alongside a significant and progressive increase of both SNR and CNR (p < 0.001). The use of incremental levels of DLIR was associated with a significant decrease of Agatston CAC score and CAC volume (p < 0.001), while mass score remained unchanged when compared to FBP (p = 0.232). The underestimation of Agatston CAC led to a CAC-DRS misclassification rate of 8%.ConclusionDLIR systematically underestimates Agatston CAC score. Therefore, DLIR should be used cautiously for cardiovascular risk assessment.Key Points• In coronary artery calcium imaging, the implementation of deep learning image reconstructions improves image quality, by decreasing the level of image noise.• Deep learning image reconstructions systematically underestimate Agatston coronary artery calcium score.• Deep learning image reconstructions should be used cautiously in clinical routine to measure Agatston coronary artery calcium score for cardiovascular risk assessment.

Modelling premature cardiac aging with induced pluripotent stem cells from a hutchinson-gilford Progeria Syndrome patient.

Hutchinson-Gilford Progeria Syndrome (HGPS) is a rare genetic disorder that causes accelerated aging and a high risk of cardiovascular complications. However, the underlying mechanisms of cardiac complications of this syndrome are not fully understood. This study modeled HGPS using cardiomyocytes (CM) derived from induced pluripotent stem cells (iPSC) derived from a patient with HGPS and characterized the biophysical, morphological, and molecular changes found in these CM compared to CM derived from a healthy donor. Electrophysiological recordings suggest that the HGPS-CM was functional and had normal electrophysiological properties. Electron tomography showed nuclear morphology alteration, and the 3D reconstruction of electron tomography images suggests structural abnormalities in HGPS-CM mitochondria, however, there was no difference in mitochondrial content as measured by Mitotracker. Immunofluorescence indicates nuclear morphological alteration and confirms the presence of Troponin T. Telomere length was measured using qRT-PCR, and no difference was found in the CM from HGPS when compared to the control. Proteomic analysis was carried out in a high-resolution system using Liquid Chromatography Tandem Mass Spectrometry (LC-MS/MS). The proteomics data show distinct group separations and protein expression differences between HGPS and control-CM, highlighting changes in ribosomal, TCA cycle, and amino acid biosynthesis, among other modifications. Our findings show that iPSC-derived cardiomyocytes from a Progeria Syndrome patient have significant changes in mitochondrial morphology and protein expression, implying novel mechanisms underlying premature cardiac aging.