A Preconditioning Approach To Optimizing Sensing Matrix For Improved Compressed Sensing CT Reconstruction

Compressed sensing (CS) image reconstruction in CT suffers from the drawbacks such as 1) appearance of staircase artifacts and 2) loss in image textures and smooth intensity changes. These drawbacks stem from the fact that CS is based on approximating the image by a piecewise-constant function. To overcome this drawback, we have already proposed a framework to improve image quality in CS using deep learning. In this framework, FBP reconstructed image and CS (TV or Nonlocal TV) reconstructed image are inputted to CNN with two input channels and single output channel, and a final reconstructed image is obtained by the output of CNN. Parameters (weight and bias) of CNN together with a regularization parameter of CS are estimated by minimizing an average least-squares loss function by using learning data, i.e. a set of triplet of degraded FBP reconstruction, CS reconstruction, and answer image. In this paper, this framework is extended to 3-D image reconstruction in helical cone-beam CT operated with lowdose scanning protocol. Parameters (weight and bias) of CNN together with a regularization parameter of CS are estimated by minimizing an average least-squares loss function by using learning data, i.e. a set of triplet of degraded FBP reconstruction, CS reconstruction, and answer image. In this paper, this framework was extended to 3-D image reconstruction in helical cone-beam CT operated with lowdose scanning protocol. The extension was done in the following way. First, we prepare N different 2-D denoising CNN (CNN₁, CNN₂, . . . , CNNN ) dependent on the slice position n. Each slice of the short-scan FDK reconstruction without denoising y_i and with 3-D TV (or Nonlocal TV) denoising z_i are inputted to CNNn with the closest slice index n, which yields a corresponding output image for each slice x_i . The final reconstructed image is obtained by stacking every slice x_i (i = 1, 2, . . . , I).

This paper compares four different minimization approaches for iterative reconstruction in CT:(1) iterative coordinate descent approach (ICD), (2) conjugate gradient approach (CG), (3) separable parabolic surrogate approach with ordered subsets (OS), and (4) convergent ordered subsets approach (COS). In addition to showing that all approaches result in the same final image, the paper gives an indication of the number of iterations and time to convergence for the studied approaches

Image reconstruction from few-view CT is of interest because of the potential to reduce scanning time and radiation dose. The challenge of few-view CT for image reconstruction is essentially a problem of interpolation from under-sampled data. Recently, a new algorithm for inverting the Fourier transform from under-sampled data has been developed by Candes et al. <i>IEEE Trans. Inf. Theory</i> , <b>52</b> 489 (2006). This algorithm can be directly applied to image reconstruction in 2D parallel-beam CT because of the central slice theorem. This article presents a discussion of the new algorithm, showing examples for different degrees of under-sampling.

Conventional CT reconstruction has been effectively addressed by analytic and iterative algorithms. Constraints on radiation dose, scanning range, and acquisition time in specific settings can render CT reconstruction ill-posed, invalidating these conventional approaches. To address these ill-posed CT reconstruction challenges, numerous task-specific deep learning solutions have emerged, demonstrating significant improvements in image quality. However, these methods are tied to particular degradation patterns and generalize poorly across degradations. In response, we propose a general-purpose reconstruction algorithm for ill-posed CT that comprehensively harnesses the potential of deep generative priors. Specifically, we first establish deep generative priors for high-quality CT images through forward and reverse Markov chain modeling in a diffusion model. During posterior sampling, we introduce a one-shot degradation model to steer sampling toward high-quality reconstructions of arbitrarily degraded CT images. Extensive experiments across three representative ill-posed CT reconstruction scenarios — low-dose CT reconstruction, limited-angle CT reconstruction, and sparse-view CT reconstruction — verified the proposed method's superiority and generalization capability over task-specific baselines.

This Special Report summarizes the 2022, AAPM grand challenge on Truth-based CT image reconstruction. To provide an objective framework for evaluating CT reconstruction methods using virtual imaging resources consisting of a library of simulated CT projection images of a population of human models with various diseases. Two hundred unique anthropomorphic, computational models were created with varied diseases consisting of 67 emphysema, 67 lung lesions, and 66 liver lesions. The organs were modeled based on clinical CT images of real patients. The emphysematous regions were modeled using segmentations from patient CT cases in the COPDGene Phase I dataset. For the lung and liver lesion cases, 1-6 malignant lesions were created and inserted into the human models, with lesion diameters ranging from 5.6 to 21.9mm for lung lesions and 3.9 to 14.9mm for liver lesions. The contrast defined between the liver lesions and liver parenchyma was 82±12 HU, ranging from 50 to 110 HU. Similarly, the contrast between the lung lesions and the lung parenchyma was defined as 781±11 HU, ranging from 725 to 805 HU. For the emphysematous regions, the defined HU values were -950±17 HU ranging from -918 to -979 HU. The developed human models were imaged with a validated CT simulator. The resulting CT sinograms were shared with the participants. The participants reconstructed CT images from the sinograms and sent back their reconstructed images. The reconstructed images were then scored by comparing the results against the corresponding ground truth values. The scores included both task-generic (root mean square error [RMSE] and structural similarity matrix [SSIM]), and task-specific (detectability index [d'] and lesion volume accuracy) metrics. For the cases with multiple lesions, the measured metric was averaged across all the lesions. To combine the metrics with each other, each metric was normalized to a range of 0 to 1 per disease type, with "0" and "1" being the worst and best measured values across all cases of the disease type for all received reconstructions. The True-CT challenge attracted 52 participants, out of which 5 successfully completed the challenge and submitted the requested 200 reconstructions. Across all participants and disease types, SSIM absolute values ranged from 0.22 to 0.90, RMSE from 77.6 to 490.5 HU, d' from 0.1 to 64.6, and volume accuracy ranged from 1.2 to 753.1mm3. The overall scores demonstrated that participant "A" had the best performance in all categories, except for the metrics of d' for lung lesions and RMSE for liver lesions. Participant "A" had an average normalized score of 0.41±0.22, 0.48±0.32, and 0.42±0.33 for the emphysema, lung lesion, and liver lesion cases, respectively. The True-CT challenge successfully enabled objective assessment of CT reconstructions with the unique advantage of access to a diverse population of diseased human models with known ground truth. This study highlights the significant potential of virtual imaging trials in objective assessment of medical imaging technologies.

BackgroundSPECT-CT using radiolabeled phosphonates is considered a standard for assessing bone metabolism (e.g., in patients with osteoarthritis of knee joints). However, SPECT can be influenced by metal artifacts in CT caused by endoprostheses affecting attenuation correction. The current study examined the effects of metal artifacts in CT of a specific endoprosthesis design on quantitative hybrid SPECT-CT imaging.The implant was positioned inside a phantom homogenously filled with activity (955 MBq 99mTc). CT imaging was performed for different X-ray tube currents (I = 10, 40, 125 mA) and table pitches (p = 0.562 and 1.375). X-ray tube voltage (U = 120 kVp) and primary collimation (16 × 0.625 mm) were kept constant for all scans. The CT reconstruction was performed with five different reconstruction kernels (slice thickness, 1.25 mm and 3.75 mm, each 512 × 512 matrix). Effects from metal artifacts were analyzed for different CT scans and reconstruction protocols. ROI analysis of CT and SPECT data was performed for two slice positions/volumes representing the typical locations for target structures relative to the prosthesis (e.g., femur and tibia). A reference region (homogenous activity concentration without influence from metal artifacts) was analyzed for comparison.ResultsSignificant effects caused by CT metal artifacts on attenuation-corrected SPECT were observed for the different slice positions, reconstructed slice thicknesses of CT data, and pitch and CT-reconstruction kernels used (all, p < 0.0001). Based on the optimization, a set of three protocols was identified minimizing the effect of CT metal artifacts on SPECT data. Regarding the reference region, the activity concentration in the anatomically correlated volume was underestimated by 8.9–10.1%. A slight inhomogeneity of the reconstructed activity concentration was detected inside the regions with a median up to 0.81% (p < 0.0001). Using an X-ray tube current of 40 mA showed the best result, balancing quantification and CT exposure.ConclusionThe results of this study demonstrate the need for the evaluation of SPECT-CT protocols in prosthesis imaging. Phantom experiments demonstrated the possibility for quantitative SPECT-CT of bone turnover in a specific prosthesis design. Meanwhile, a systematic bias caused by metal implants on quantitative SPECT data has to be considered.

We report a case of chondrodysplasia punctata tibia-metacarpal type (CDP-TM) that was diagnosed prenatally using multidetector CT (MDCT) with three-dimensional (3-D) CT reconstructions. Prenatal US had shown severe thoracic hypoplasia and rhizomelic shortening of the limbs, raising the suspicion of thanatophoric dysplasia. However, MDCT showed punctate calcifications in the epiphyseal cartilage of the humeri and femora, carpal bones, and paravertebral region. On 3-D CT, the tibiae were much shorter than the fibulae, the humeri were very short and bowed, and severe platyspondyly was evident. These findings led to the diagnosis of CDP-TM. The diagnosis was confirmed on postnatal radiographs. Prenatal MDCT with 3-D images may make a useful contribution to prenatal diagnosis in selected fetuses with severe skeletal dysplasia.

To provide a proof of concept validation of a novel 4D cone-beam CT (4DCBCT) reconstruction algorithm and to determine the best methods to train and optimize the algorithm. The algorithm animates a patient fan-beam CT (FBCT) with a patient specific parametric motion model in order to generate a time series of deformed CTs (the reconstructed 4DCBCT) that track the motion of the patient anatomy on a voxel by voxel scale. The motion model is constrained by requiring that projections cast through the deformed CT time series match the projections of the raw patient 4DCBCT. The motion model uses a basis of eigenvectors that are generated via principal component analysis (PCA) of a training set of displacement vector fields (DVFs) that approximate patient motion. The eigenvectors are weighted by a parameterized function of the patient breathing trace recorded during 4DCBCT. The algorithm is demonstrated and tested via numerical simulation. The algorithm is shown to produce accurate reconstruction results for the most complicated simulated motion, in which voxels move with a pseudo-periodic pattern and relative phase shifts exist between voxels. The tests show that principal component eigenvectors trained on DVFs from a novel 2D/3D registration method give substantially better results than eigenvectors trained on DVFs obtained by conventionally registering 4DCBCT phases reconstructed via filtered backprojection. Proof of concept testing has validated the 4DCBCT reconstruction approach for the types of simulated data considered. In addition, the authors found the 2D/3D registration approach to be our best choice for generating the DVF training set, and the Nelder-Mead simplex algorithm the most robust optimization routine.

The aim of this study is to compare the diagnostic value of two-dimensional (2D) CT and three-dimensional (3D) CT reconstruction techniques in detecting maxillofacial fractures in patients at Mzuzu Central Hospital (MCH). 67 maxillofacial trauma patients admitted to Mzuzu Central Hospital from Jan to Sep 2024 underwent multi-slice spiral CT (MSCT) scanning. Images were post-processed using 2D and 3D reconstruction techniques. Clinical and radiological data were collected from the patients, and a comparative analysis of the results from the two reconstruction techniques was performed. In this study, 52 cases of maxillofacial fractures with a total of 83 fractures were diagnosed by 2D CT reconstruction technology, with a fracture detection rate of 77.61% (52/67). Using 3D CT reconstruction technology, 54 cases of maxillofacial fractures with a total of 91 fractures were diagnosed, and the fracture detection rate was 80.60% (54/67). Statistical analysis showed no significant difference in the detection rate of maxillofacial fractures between 2D CT and 3D CT reconstruction (χ2 = 35.945, P = 0.687). In the diagnosis of zygomatic fractures, nasal fractures, and upper and lower jaw fractures, 3D CT reconstruction images have obvious advantages over 2D CT in displaying fracture displacement and fracture line course. However, for the display of comminuted fractures combined with sphenoid and ethmoid fractures, the cross-sectional images of 2D CT show higher superiority. 2D CT reconstruction is a basic diagnostic tool for maxillofacial fractures. 3D reconstruction, with high detection and multi-angle visualization, offers valuable imaging for clinical decision-making, aiding in surgery planning. A combined approach, leveraging the strengths of both modalities, is pivotal for comprehensive assessment and management of maxillofacial trauma.

The advancement of x-ray CT into the domains of photon counting spectral imaging and dynamic cardiac and perfusion imaging has created many new challenges and opportunities for clinicians and researchers. To address challenges such as dose constraints and scanning times while capitalizing on opportunities such as multi-contrast imaging and low-dose coronary angiography, these multi-channel imaging applications require a new generation of CT reconstruction tools. These new tools should exploit the relationships between imaging channels during reconstruction to set new image quality standards while serving as a platform for direct translation between the preclinical and clinical domains. We outline and demonstrate a new Multi-Channel Reconstruction (MCR) Toolkit for GPU-based analytical and iterative reconstruction of preclinical and clinical multi-energy and dynamic x-ray CT data. To promote open science, open-source distribution of the Toolkit will coincide with the release of this publication (GPL v3; gitlab.oit.duke.edu/dpc18/mcr-toolkit-public). The MCR Toolkit source code is implemented in C/C++ and NVIDIA's CUDA GPU programming interface, with scripting support from MATLAB and Python. The Toolkit implements matched, separable footprint CT reconstruction operators for projection and backprojection in two geometries: planar, cone-beam CT (CBCT) and 3rd generation, cylindrical multi-detector row CT (MDCT). Analytical reconstruction is performed using filtered backprojection (FBP) for circular CBCT, weighted FBP (WFBP) for helical CBCT, and cone-parallel projection rebinning followed by WFBP for MDCT. Arbitrary combinations of energy and temporal channels are iteratively reconstructed under a generalized multi-channel signal model for joint reconstruction. We solve this generalized model algebraically using the split Bregman optimization method and the BiCGSTAB(l) linear solver interchangeably for both CBCT and MDCT data. Rank-sparse kernel regression (RSKR) and patch-based singular value thresholding (pSVT) are used to regularize the energy and time dimensions, respectively. Under a Gaussian noise model, regularization parameters are estimated automatically from the input data, dramatically reducing algorithm complexity for end users. Multi-GPU parallelization of the reconstruction operators is supported to manage reconstruction times. Denoising with RSKR and pSVT and post-reconstruction material decomposition are illustrated with preclinical and clinical cardiac photon-counting (PC)CT data. A digital MOBY mouse phantom with cardiac motion is used to illustrate single energy (SE), multi-energy (ME), time resolved (TR), and combined multi-energy and time-resolved (METR) helical, CBCT reconstruction. A fixed set of projection data is used across all reconstruction cases to demonstrate the Toolkit's robustness to increasing data dimensionality. Identical reconstruction code is applied to in vivo cardiac PCCT data acquired in a mouse model of atherosclerosis (METR). Clinical cardiac CT reconstruction is illustrated using the XCAT phantom and the DukeSim CT simulator, while dual-source, dual-energy CT reconstruction is illustrated for data acquired with a Siemens Flash scanner. Benchmarking results with NVIDIA RTX 8000 GPU hardware demonstrate 61%-99% efficiency in scaling computation from one to four GPUs for these reconstruction problems. The MCR Toolkit provides a robust solution for temporal and spectral x-ray CT reconstruction problems and was built from the ground up to facilitate translation of CT research and development between preclinical and clinical applications.

To compare the accuracy of three-dimensional reconstruction of cervical CT and ultrasound for estimating residual thyroid volume. We performed a retrospective analysis of 17 patients with 21 residual thyroid glands undergoing thyroidectomy surgery between February, 2017 and March, 2018 in our department. We compared the residual thyroid volume in preoperative ultrasound with the intraoperative measurement and the volume measured using threedimensional CT reconstruction before surgery. The maximum vertical and anterioposterior diameters of the residual thyroid measured by preoperative ultrasound differed significantly from the volume data measured intraoperatively (P &lt; 0.05), but the difference in the maximum left-right diameters was not statistically significant (P&gt;0.05). The maximum vertical, leftright, and anteroposterior diameters estimated by three-dimensional reconstruction of cervical CT was all similar with those measured intraoperatively (P&gt;0.05). Compared with ultrasound examination, three-dimensional reconstruction of neck CT is more accurate for estimating the residual thyroid volume and provides more reliable evidence for clinical calculation of postoperative I131 dose for thyroid cancer.

Limiting scan views is an efficient way to reduce radiation doses in the cone-beam computed tomography (CBCT) examinations, which unfortunately degrades the reconstructed images. Some methods on the framework of the generative adversarial network (GAN) were developed to improve low-dose CT images after CT reconstruction from the limited-view projections. However, no GAN-based methods were devoted to restoring missing CBCT projections in the sinogram domain before CT reconstruction. To avoid the trade-off between radiation dose and image quality, we propose a limited-view CBCT reconstruction method in the sinogram domain, instead of the image domain. First, this method slices the 3D CBCT projections into multiple 2D pieces. Then, an adversarial autoencoder network is trained to estimate the missing parts of these 2D pieces. To improve the prediction, we apply a joint loss function, including reconstruction loss and adversarial loss to the network. When the new limited-view 3D CBCT projections are acquired, the proposed method uses the trained adversarial autoencoder network to generate the missing parts of the 2D pieces sliced from the current 3D CBCT projections. Then, stacking the completed 2D pieces in order yields full-view 3D CBCT projections. Finally, we reconstruct the CT images from the full-view 3D CBCT projections by using the Feldkamp, Davis, and Kress algorithm. The experiments validate that our method performs well in the prediction of unknown projections and CT reconstruction and are less vulnerable to the number of unknown projections than other methods.

Purpose: Iterative reconstruction via total variation (TV) minimization has demonstrated great successes in accurate CT imaging from under‐sampled projections. When projections are further reduced, over‐smoothing artifacts appear in the current reconstruction especially around the structure boundaries. We propose a practical algorithm to improve TV‐minimization based CT reconstruction on very few projection data. Methods: The L‐0 norm approach is more desirable from the perfective of further reducing the projection views. To overcome the computational difficulty of the non‐convex optimization of the L‐0 norm, we implement an adaptive weighting scheme to approximate the solution via a series of TV minimizations for practical use in CT reconstruction. The weight on TV is initialized as uniform ones, and is adaptively changed based on the gradient of the reconstructed image from the previous iteration. The iteration stops when a small difference between the weighted TV values is observed on two consecutive reconstructed images. Results: On the digital Shepp‐Logan phantom, the proposed method reduces reconstruction errors in the conventional TV minimization from 7.3% to 1.4% with 20 projections, and from 25% to 6.5% with 15 projections. With 20 projections on the Catphan600 phantom, our method reduces contrast errors in the ROIs from 45 HU to 6 HU. The rise width at the object edges in the image is also reduced by 40%, showing a substantial image resolution improvement. Conclusion: By adaptively reweighting TV in iterative CT reconstruction, we successfully further reduce the projection number for the same or better image quality. The technique is attractive in the applications of CT reconstruction on a small size of projection data.

Background Results of recent phantom studies show that variation in CT acquisition parameters and reconstruction techniques may make radiomic features largely nonreproduceable and of limited use for prognostic clinical studies. Purpose To investigate the effect of CT radiation dose and reconstruction settings on the reproducibility of radiomic features, as well as to identify correction factors for mitigating these sources of variability. Materials and Methods This was a secondary analysis of a prospective study of metastatic liver lesions in patients who underwent staging with single-energy dual-source contrast material-enhanced staging CT between September 2011 and April 2012. Technique parameters were altered, resulting in 28 CT data sets per patient that included different dose levels, section thicknesses, kernels, and reconstruction algorithm settings. By using a training data set (n = 76), reproducible intensity, shape, and texture radiomic features (reproducibility threshold, R2 ≥ 0.95) were selected and correction factors were calculated by using a linear model to convert each radiomic feature to its estimated value in a reference technique. By using a test data set (n = 75), the reproducibility of hierarchical clustering based on 106 radiomic features measured with different CT techniques was assessed. Results Data in 78 patients (mean age, 60 years ± 10; 33 women) with 151 liver lesions were included. The percentage of radiomic features deemed reproducible for any variation of the different technical parameters was 11% (12 of 106). Of all technical parameters, reconstructed section thickness had the largest impact on the reproducibility of radiomic features (12.3% [13 of 106]) if only one technical parameter was changed while all other technical parameters were kept constant. The results of the hierarchical cluster analysis showed improved clustering reproducibility when reproducible radiomic features with dedicated correction factors were used (ρ = 0.39-0.71 vs ρ = 0.14-0.47). Conclusion Most radiomic features are highly affected by CT acquisition and reconstruction settings, to the point of being nonreproducible. Selecting reproducible radiomic features along with study-specific correction factors offers improved clustering reproducibility. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Sosna in this issue.

We emphasise the importance of high-resolution CT with reconstruction in the demonstration of submandibular gland (SMG) sialolithiasis and its role in monitoring treatment. We studied 76 patients with swollen and tender SMG, some with fever. They underwent conventional radiography, sonography (US) and high-resolution CT with reconstructions. Conventional radiographs demonstrated single stones in 29 patients. Axial CT, before reconstructions, demonstrated single stones in 63 patients and multiple stones in another 5. Following CT reconstructions, multiple stones were demonstrated in 37 patients. On US stones were diagnosed in only 33 patients, and multiple stones in only 1. All 68 patients with stones shown on imaging and 2 without stones underwent surgery, with good clinical results. Total removal of the SMG and its duct was performed in patients with multiple stones, chronic inflammatory changes in the SMG, or a solitary stone in the SMG or deep in the duct. A small incision for removal of a solitary stone in the distal aspect of Wharton's duct was performed in 15 patients, with excellent clinical results. Another 14 patients with multiple salivary gland stones, diagnosed on CT reconstructions, did not improve following this procedure and needed further surgery; clinical improvement occurred following excision of the SMG and Wharton's duct. Histological examination in all of these confirmed the presence of additional stones. Conservative anti-inflammatory treatment was recommended for 6 patients in whom CT reconstructions did not demonstrate stones.