Weakly Supervised Limited-Angle CT Reconstruction Using Multiple Projection Geometries

Purpose: To investigate dose distributions in the prostate and surrounding tissues for I–125 brachytherapy using Monte Carlo (MC) calculations with patient‐specific phantoms derived from CT images. We explore the effectiveness of different techniques to mitigate streaking artifacts due to brachytherapy sources in post‐implant CT images. Methods: Nine patients (45–87 sources) are considered. Streaking artifacts in post‐implant CT images are mitigated using various metallic artifact reduction (MAR) techniques: raw sinogram, fan beam virtual sinogram and 3D median filter. Segmented structures (CTV, rectum and bladder) guide the assignment of tissues (air, muscle, prostate, calcification, average tissue and bone) to develop patient‐specific MC phantoms. The EGSnrc user‐code Brachydose and GEANT4 user‐code ALGEBRA are employed for MC dose calculations using patient‐specific phantoms derived from CT images. Dose distributions generated with patient models derived using different MAR techniques and different calculation methods (MC, TG‐43) are compared directly, and using recommended dose metrics. Results: Application of each MAR technique to patient CT data results in comparable mitigation of streaking artifacts within the treatment volume in CT images. MC calculations based on uncorrected CT data Result in high dose spikes (>200%) in the treatment volume and lower doses (40–70%) in surrounding tissues compared to MC calculations with corrected phantoms. Dose distributions within the prostate from MC simulations using phantoms generated with different MAR techniques are comparable to each other but differ from TG‐43 calculations with significant (10%+) local differences. Conclusion: Mitigation of streaking artifacts in CT images is necessary for patient‐specific MC dose calculations; however, dose distributions and clinical metrics in the target and OAR are insensitive to the particular MAR technique applied. Differences between MC and TG‐43 dose distributions within the target volume and surrounding organs for I‐125 prostate brachytherapy underline the importance of patient‐specific model‐based dose calculations for treatment planning and evaluation. NSERC, CCSRI, The Canada Research Chairs Program

Geometric dependence of image quality in digital tomosynthesis: Simulations of X-ray source trajectories and scan angles

We developed a high-resolution CT system using synchrotron radiation. The system acquires high-quality CT images because the synchrotron CT has no beam-hardening effect that results from using monochromatic X-ray. It is said that X-ray scattering has no effects on the synchrotron CT images because a detector is set apart from a subject. However, the images have a disadvantage in non-uniformity caused by scattered X-rays owing to an area X-ray beam and a two-dimensional detector. The X-ray scattering decreased the accuracy of the CT number and increased the number of artifacts in CT images. We developed an accurate process for correcting scattered X-rays for the synchrotron CT. The correction process was applied for the synchrotron CT at 35-keV radiation. When a 10-mm-diameter 20%-hydroxyapatite homogeneous column phantom was used, the scattered X-rays led to a reduction of about 45% in the pixel values of CT images. With the correction, the pixel values were almost completely recovered.

The goal of this work was to investigate the effects of MRI surface coils on attenuation-corrected PET emission data. The authors studied the cases where either an MRI or a CT scan would be used to provide PET attenuation correction (AC). Combined MR/PET scanners that use the MRI for PET AC (MR-AC) face the challenge of absent surface coils in MR images and thus cannot directly account for attenuation in the coils. Combining MR and PET images could be achieved by transporting the subject on a stereotactically registered table between independent MRI and PET scanners. In this case, conventional PET CT-AC methods could be used. A challenge here is that high atomic number materials within MR coils cause artifacts in CT images and CT based AC is typically not validated for coil materials. The authors evaluated PET artifacts when MR coils were absent from AC data (MR-AC), or when coil attenuation was measured by CT scanning (CT-AC). They scanned PET phantoms with MR surface coils on a clinical PET/CT system and used CT-AC to reconstruct PET data. The authors then omitted the coil from the CT-AC image to mimic the MR-AC scenario. Images were acquired using cylinder and anthropomorphic phantoms. They evaluated and compared the following five scenarios: (1) A uniform cylinder phantom and head coil scanned and reconstructed using CT-AC; (2) similar emission data (with head coil present) were reconstructed without the head coil in the AC data; (3) the same cylinder scanned without the head coil present (reference scan); (4) a PET torso phantom with a full MR torso coil present in both PET and CT; (5) only half of the separable torso coil present in the PET/CT acquisition. The authors also performed analytic simulations of the first three scenarios. Streak artifacts were present in CT images containing MR surface coils due to metal components. These artifacts persisted after the CT images were converted for PET AC. The artifacts were significantly reduced when half of the separable coil was removed during the scan. CT scans tended to over-estimate the linear attenuation coefficient (micro) of the metal components when using conventional methods for converting from CT number to micro(511 keV). Artifacts were visible outside the phantom in some of the PET emission images, corresponding to the MRI coil geometry. However, only subtle artifacts were apparent in the emission images inside the phantoms. On the other hand, the PET emission image quantitative accuracy was significantly affected: the activity was underestimated by 19% when AC did not include the head coil, and overestimated by 28% when the CT-AC included the head coil. The presence of MR coils during PET or PET/CT scanning can cause subtle artifacts and potentially important quantification errors. Alternative CT techniques that mitigate artifacts should be used to improve AC accuracy. When possible, removing segments of an MR coil prior to the PET/CT exam is recommended. Further, MR coils could be redesigned to reduce artifacts by rearranging placement of the most attenuating materials.

The presence of high density objects leads to significant artifacts in CT images. These artifacts impact the quantitative as well as qualitative accuracy of CT images. These artifacts are caused due to factors such as beam hardening, scatter, photon starvation. The ramp filter prior to standard back-projection enhances some of the artifacts. The artifacts can be reduced by using better data acquisition such as dual energy imaging, higher kVp imaging. Several software based techniques have been proposed to reduce metal artifacts that can be classified into model-based algorithms and sinogram in-painting methods. We propose an improved metal artifact correction algorithm that belongs to the category of sinogram inpainting. In the prior art, the prior images used to generate the inpainted data is created by segmenting the original or the first pass metal artifact reduced (MAR) images. We propose a multi-band filter design to generate the prior image. The original image and the first pass MAR image possess complimentary information and are combined using a multi-band filter. The combined image is then segmented to generate the final prior image. It is shown that the new approach leads to a prior that is more consistent to the original image compared to the conventional prior and hence improved inpainted data. The proposed approach is demonstrated to be superior to the conventional approach using clinical datasets. We further compare two different inpainting algorithms to replace the original corrupted sinogram samples with the forward projection of the prior, also defined as the prior data. The first approach is based on the linear baseline shift algorithm, while in the second approach the replacement step used in the normalized metal artifact correction algorithm (NMAR) is used. Both the approaches are validated using both phantom and clinical data and is demonstrated to be superior to standard interpolation based techniques.

To evaluate the performance and reliability of the single-energy metal artifact reduction (SEMAR) algorithm in patients with different orthopedic hardware at the hips. A total of 153 patients with hip instrumentation who had undergone CT with adaptive iterative dose reduction (AIDR) 3D and SEMAR algorithms between February 2015 and October 2019 were included retrospectively. Patients were divided into 5 groups by the hardware type. Two readers with 21 and 13years of experience blindly reviewed all image sets and graded the extent of artifacts and imaging quality using 5-point scales. To evaluate reliability, the mean densities and image noises were measured at the urinary bladder, veins, and fat in images with artifacts and the reference images. No significant differences were found in the mean densities of the urinary bladder, veins, and fat between the SEMAR images with artifacts (7.57 ± 9.49, 40.29 ± 23.07, 86.78 ± 38.34) and the reference images (7.77 ± 6.2, 40.27 ± 8.66, 89.10 ± 20.70) (P = .860, .994, .392). Image noises of the urinary bladder in the SEMAR images with artifacts (14.25 ± 4.50) and the SEMAR reference images (9.69 ± 1.29) were significantly higher than those in the AIDR 3D reference images (9.11 ± 1.12) (P < .001; P < .001). All AIDR 3D images were non-diagnostic (overall quality ≤ 3) and less than a quarter of the SEMAR images were non-diagnostic (16.7-23.7%), mainly in patients with prostheses [reader 1: 91.7% (22/24); reader 2: 92.6% (25/27)]. The SEMAR algorithm significantly reduces metal artifacts in CT images, more in patients with internal fixations than in patients with prostheses, and provides reliable attenuation of soft tissues.

Three dimensional (3D) models of deceased and injured people in combination with 3D scans of injury causing objects can assist forensic investigations in reconstructing event scenes. Medical imaging techniques, such as post-mortem computed tomography (PMCT) and post-mortem magnetic resonance imaging (PMMR), have been successfully applied to forensic investigations and can add beneficial value to standard autopsy examinations. These imaging modalities can be helpful for 3D reconstructions, especially when internal findings, such as bone fractures, organ damage and internal bleeding, are relevant for the investigation. However, none of these techniques can adequately visualize pattern injuries, such as boot prints and bite marks, or any type of blunt force trauma that forms distinct discolorations on the body’s surface. This is why 3D surface imaging techniques have been introduced to the forensic community. Unfortunately, many commercially available optical scanning systems are cost intensive, time consuming and can only be applied before or after a CT scan has been performed. In this article, we present a mobile, multi-camera rig based on close-range photogrammetry that is inexpensive, fast in acquisition time and can be combined with automated CT scanning protocols. The multi-camera setup comprises seven digital single-lens reflex (DSLR) cameras that are mounted on a mobile frame. Each camera is equipped with a remote control that can trigger the shutter release of all cameras simultaneously. In combination with a medical CT scanner, image acquisition of the multi camera setup can be included into an automated CT scanning procedure. In our preliminary study, textured 3D models of one side of the body were created in less than 15 minutes. The photo acquisition time combined with the modified CT scanning protocols lasted 3:34 minutes whereas the subsequent computation of a textured 3D model based on a low resolution mesh lasted 10:55 minutes. The mobile, multi-camera setup can also be used manually in combination with examination couches, lifting carts and autopsy tables. Finally, the system is not limited to post-mortem investigations but can also be applied to living people and may be used in clinical settings.

Tillage effects on topsoil structural quality assessed using X-ray CT, soil cores and visual soil evaluation

At least three gantry geometries are widely used in medical CT scanners: (1) rotate-translate, (2) rotating detectors, (3) stationary detectors. There are significant geometrical differences between these designs, especially regarding (a) the region of space scanned by any given detector and (b) the sample density of which scan the patient. It is imperative to distinguish between and rays in analyzing this situation. In particular, views are defined by the x-ray source in type 2 and by the detector in type 3 gantries. It is known that ray dependent errors are generally much more important than view dependent errors. It is shown that spatial resolution is primarily limited by the spacing between in any view, while the number of ray samples per beam width determines the extent of aliasing artifacts. Rotating detector gantries are especially susceptible to aliasing effects. It is shown that aliasing effects can distort the point spread function in a way that is highly dependent on the position of the point in the scanned field. Such effects can cause anomalies in the MTF functions as derived from points in machines with significant aliasing problems.

To suppress the streak artifacts in images reconstructed from sparse-view projections in computed tomography (CT), a residual, attention-based, dense UNet (RAD-UNet) deep network is proposed to achieve accurate sparse reconstruction. The filtered back projection (FBP) algorithm is used to reconstruct the CT image with streak artifacts from sparse-view projections. Then, the image is processed by the RAD-UNet to suppress streak artifacts and obtain high-quality CT image. Those images with streak artifacts are used as the input of the RAD-UNet, and the output-label images are the corresponding high-quality images. Through training via the large-scale training data, the RAD-UNet can obtain the capability of suppressing streak artifacts. This network combines residual connection, attention mechanism, dense connection and perceptual loss. This network can improve the nonlinear fitting capability and the performance of suppressing streak artifacts. The experimental results show that the RAD-UNet can improve the reconstruction accuracy compared with three existing representative deep networks. It may not only suppress streak artifacts but also better preserve image details. The proposed networks may be readily applied to other image processing tasks including image denoising, image deblurring, and image super-resolution.

Photon-counting computed tomography (Photon counting CT) utilizes photon-counting detectors to precisely count incident photons and measure their energy. These detectors, compared to traditional energy integration detectors, provide better image contrast and material differentiation. However, Photon counting CT tends to show more noticeable ring artifacts due to limited photon counts and detector response variations, unlike conventional spiral CT. To comprehensively address this issue, we propose a novel feature shared multi-decoder network (FSMDN) that utilizes complementary learning to suppress ring artifacts in Photon counting CT images. Specifically, we employ a feature-sharing encoder to extract context and ring artifact features, facilitating effective feature sharing. These shared features are also independently processed by separate decoders dedicated to the context and ring artifact channels, working in parallel. Through complementary learning, this approach achieves superior performance in terms of artifact suppression while preserving tissue details. We conducted numerous experiments on Photon counting CT images with three-intensity ring artifacts. Both qualitative and quantitative results demonstrate that our network model performs exceptionally well in correcting ring artifacts at different levels while exhibiting superior stability and robustness compared to the comparison methods. In this paper, we have introduced a novel deep learning network designed to mitigate ring artifacts in Photon counting CT images. The results illustrate the viability and efficacy of our proposed network model as a new deep learning-based method for suppressing ring artifacts.

Sparse-view computed tomography (CT) has been attracting attention for its reduced radiation dose and scanning time. However, analytical image reconstruction methods suffer from streak artifacts due to insufficient projection views. Recently, various deep learning-based methods have been developed to solve this ill-posed inverse problem. Despite their promising results, they are easily overfitted to the training data, showing limited generalizability to unseen systems and patients. In this work, we propose a novel streak artifact reduction algorithm that provides a system- and patient-specificsolution. Motivated by the fact that streak artifacts are deterministic errors, we regenerate the same artifacts from a prior CT image under the same system geometry. This prior image need not be perfect but should contain patient-specific information and be consistent with full-view projection data for accurate regeneration of the artifacts. To this end, we use a coordinate-based neural representation that often causes image blur but can greatly suppress the streak artifacts while having multiview consistency. By employing techniques in neural radiance fields originally proposed for scene representations, the neural representation is optimized to the measured sparse-view projection data via self-supervised learning. Then, we subtract the regenerated artifacts from the analytically reconstructed original image to obtain the final correctedimage. To validate the proposed method, we used simulated data of extended cardiac-torso phantoms and the 2016 NIH-AAPM-Mayo Clinic Low-Dose CT Grand Challenge and experimental data of physical pediatric and head phantoms. The performance of the proposed method was compared with a total variation-based iterative reconstruction method, naive application of the neural representation, and a convolutional neural network-based method. In visual inspection, it was observed that the small anatomical features were best preserved by the proposed method. The proposed method also achieved the best scores in the visual information fidelity, modulation transfer function, and lung nodulesegmentation. The results on both simulated and experimental data suggest that the proposed method can effectively reduce the streak artifacts while preserving small anatomical structures that are easily blurred or replaced with misleading features by the existing methods. Since the proposed method does not require any additional training datasets, it would be useful in clinical practice where the large datasets cannot becollected.

CT images often have some ring artefacts caused by the inconsistent response of detector pixels and other factors. Removing ring artefacts effectively without impairing the image quality is still a hard problem although some methods have been proposed. Given that total variation minimisation is widely used for image denoising, a unidirectional relative variation model is proposed and performs it on the reconstructed images transformed into polar coordinates to reduce the artefacts. The proposed method is tested on both simulated data and real CT images, and the qualitative and quantitative results show that it is effective.

PurposeAs computed tomography (CT) imaging is the most commonly used modality for treatment planning in radiation therapy, metal artifacts in the planning CT images may complicate the target delineation and reduce the dose calculation accuracy. Although current CT scanners do provide certain correction steps, it is a common understanding that there is not a universal solution yet to the metal artifact reduction (MAR) in general. Particularly noting the importance of MAR for radiation treatment planning, we propose a novel MAR method in this work that recruits an additional tilted CT scan and synthesizes nearly metal‐artifact‐free CT images.MethodsThe proposed method is based on the facts that the most pronounced metal artifacts in CT images show up along the x‐ray beam direction traversing multiple metallic objects and that a tilted CT scan can provide complementary information free of such metal artifacts in the earlier scan. Although the tilted CT scan would contain its own metal artifacts in the images, the artifacts may manifest in a different fashion leaving a chance to concatenate the two CT images with the metal artifacts much suppressed. We developed an image processing technique that uses the structural similarity (SSIM) for suppressing the metal artifacts. On top of the additional scan, we proposed to use an existing MAR method for each scan if necessary to further suppress the metal artifacts.ResultsThe proposed method was validated by a simulation study using the pelvic region of an XCAT numerical phantom and also by an experimental study using the head part of the Rando phantom. The proposed method was found to effectively reduce the metal artifacts. Quantitative analyses revealed that the proposed method reduced the mean absolute percentages of the error by up to 86% and 89% in the simulation and experimental studies, respectively.ConclusionsIt was confirmed that the proposed method, using complementary information acquired from an additional tilted CT scan, can provide nearly metal‐artifact‐free images for the treatment planning.

A New Method to Identify and to Correct Circular Artifacts in X-Ray CT Images.