Artifact analysis of approximate helical cone-beam CT reconstruction algorithms

Removing ring artifacts in CBCT images via Transformer with unidirectional vertical gradient loss

BACKGROUND

Cone-beam computed tomography (CBCT), an important medical modality for disease detection and diagnosis, is currently widely used in clinical practice. However, due to the inconsistent response of CBCT detectors, the lack of proper calibration often leads to the occurrence of ring artifacts in CBCT-reconstructed images. These artifacts may affect physicians' assessment and diagnosis. Therefore, effective elimination of ring artifacts in CBCT images without degrading image quality is important.

PURPOSE

Given the pros and cons of existing methods for removing ring artifacts in CBCT images, this paper is devoted to designing a specific Transformer for this task, leveraging the global and local modeling ability of Transformer.

METHODS

We design a loss function with dual-domain information fusion for the vanilla Transformer to correct ring artifacts in CBCT images. The method operates in image domain to predict artifact-free outputs and preserve more image details. Meanwhile, we design a tailored loss function incorporating polar domain optimization to remove ring artifacts more effectively. Specifically, an unidirectional gradient loss that constrains vertical gradients in polar domain is imposed, based on the geometric prior that in polar coordinates, ring artifacts predominately affect horizontal gradients while minimally influencing vertical gradients.

RESULTS

We conduct extensive ablative and comparative experiments on CBCT/CT image set to validate the performance of the proposed method. First, four ablation experiments demonstrate the feasibility of our approach. Then, we compare our method with several classical methods and the latest state-of-the-arts, and our method achieves the highest quality of corrected images as well as the best evaluation metrics. In these experiments, 5332 CT images were used for training, and 550 CT images, and 500 real CBCT images were used for testing. The source code is available at https://github.com/shasha521/CBCT.

CONCLUSIONS

Experimental results demonstrate that our method can significantly improve the effectiveness of ring artifact correction. By capitalizing on dual-domain information fusion and a customized loss function, the improved Transformer can not only effectively remove ring artifacts in CBCT images, but also preserve the details of original images quite well.

Why do commercial CT scanners still employ traditional, filtered back-projection for image reconstruction?

Despite major advances in x-ray sources, detector arrays, gantry mechanical design and especially computer performance, one component of computed tomography (CT) scanners has remained virtually constant for the past 25 years-the reconstruction algorithm. Fundamental advances have been made in the solution of inverse problems, especially tomographic reconstruction, but these works have not been translated into clinical and related practice. The reasons are not obvious and seldom discussed. This review seeks to examine the reasons for this discrepancy and provides recommendations on how it can be resolved. We take the example of field of compressive sensing (CS), summarizing this new area of research from the eyes of practical medical physicists and explaining the disconnection between theoretical and application-oriented research. Using a few issues specific to CT, which engineers have addressed in very specific ways, we try to distill the mathematical problem underlying each of these issues with the hope of demonstrating that there are interesting mathematical problems of general importance that can result from in depth analysis of specific issues. We then sketch some unconventional CT-imaging designs that have the potential to impact on CT applications, if the link between applied mathematicians and engineers/physicists were stronger. Finally, we close with some observations on how the link could be strengthened. There is, we believe, an important opportunity to rapidly improve the performance of CT and related tomographic imaging techniques by addressing these issues.

Advances in 4D medical imaging and 4D radiation therapy

This paper reviews recent advances in 4D medical imaging (4DMI) and 4D radiation therapy (4DRT), which study, characterize, and minimize patient motion during the processes of imaging and radiotherapy. Patient motion is inevitably present in these processes, producing artifacts and uncertainties in target (lesion) identification, delineation, and localization. 4DMI includes time-resolved volumetric CT, MRI, PET, PET/CT, SPECT, and US imaging. To enhance the performance of these volumetric imaging techniques, parallel multi-detector array has been employed for acquiring image projections and the volumetric image reconstruction has been advanced from the 2D to the 3D tomography paradigm. The time information required for motion characterization in 4D imaging can be obtained either prospectively or retrospectively using respiratory gating or motion tracking techniques. The former acquires snapshot projections for reconstructing a motion-free image. The latter acquires image projections continuously with an associated timestamp indicating respiratory phases using external surrogates and sorts these projections into bins that represent different respiratory phases prior to reconstructing the cyclical series of 3D images. These methodologies generally work for all imaging modalities with variations in detailed implementation. In 4D CT imaging, both multi-slice CT (MSCT) and cone-beam CT (CBCT) are applicable in 4D imaging. In 4D MR imaging, parallel imaging with multi-coil-detectors has made 4D volumetric MRI possible. In 4D PET and SPECT, rigid and non-rigid motions can be corrected with aid of rigid and deformable registration, respectively, without suffering from low statistics due to signal binning. In 4D PET/CT and SPECT/CT, a single set of 4D images can be utilized for motion-free image creation, intrinsic registration, and attenuation correction. In 4D US, volumetric ultrasonography can be employed to monitor fetal heart beating with relatively high temporal resolution. 4DRT aims to track and compensate for target motion during radiation treatment, minimizing normal tissue injury, especially critical structures adjacent to the target, and/or maximizing radiation dose to the target. 4DRT requires 4DMI, 4D radiation treatment planning (4D RTP), and 4D radiation treatment delivery (4D RTD). Many concepts in 4DRT are borrowed, adapted and extended from existing image-guided radiation therapy (IGRT) and adaptive radiation therapy (ART). The advantage of 4DRT is its promise of sparing additional normal tissue by synchronizing the radiation beam with the moving target in real-time. 4DRT can be implemented differently depending upon how the time information is incorporated and utilized. In an ideal situation, the motion adaptive approach guided by 4D imaging should be applied to both RTP and RTD. However, until new automatic planning and motion feedback tools are developed for 4DRT, clinical implementation of ideal 4DRT will meet with limited success. However, simplified forms of 4DRT have been implemented with minor modifications of existing planning and delivery systems. The most common approach is the use of gating techniques in both imaging and treatment, so that the planned and treated target localizations are identical. In 4D planning, the use of a single planning CT image, which is representative of the statistical respiratory mean, seems preferable. In 4D delivery, on-site CBCT imaging or 3D US localization imaging for patient setup and internal fiducial markers for target motion tracking can significantly reduce the uncertainty in treatment delivery, providing improved normal tissue sparing. Most of the work on 4DRT can be regarded as a proof-of-principle and 4DRT is still in its early stage of development.

Isotropic CT examination of abdomen and pelvis diagnostic quality of reformat

RATIONALE AND OBJECTIVES

To evaluate the image quality of axial and coronal reformats obtained from isotropic resolution abdomino-pelvic computed tomography (CT) examinations.

MATERIALS AND METHODS

Thirty consecutive patients with intravenous contrast-enhanced abdomino-pelvic CT examinations (Brilliance 40, Philips Medical Systems, Cleveland, OH) were enrolled for the study. The raw data were reconstructed into two sets of source axial images: 0.9-mm slice widths with 0.45-mm reconstruction interval (isotropic resolution) and 4-mm slice widths with 3-mm reconstruction interval (anisotropic resolution: group A). Isotropic data set was reformatted into axial and coronal stacks (groups B and C, respectively) with 4-mm slice width and 3-mm interval. Three independent readers evaluated stacks A to C using a 3-point scale for resolution of hepatic vessels, edge sharpness of kidneys, respiratory motion artifact, reconstruction artifact, noise, and overall image quality.

RESULTS

There was no statistical difference among the groups A to C for vessel resolution, motion artifact, noise, and overall quality. The scores given to group C were significantly lower than those to groups A and B for reconstruction artifacts. There was no difference among groups A to C for overall impression of image quality. The interreader agreements were excellent for axial images (groups A and B) and moderate for coronal reformats.

CONCLUSION

Isotropic scanning of the abdomen and pelvis allows creation of reformats with similar image quality as similar thickness axial source images. These reformats are of sufficient quality to form the basis of clinical interpretation.

Physical evaluation of multidetector-row computed tomography (MDCT) scan methods and conditions for improvement of carbon beam distribution

To reduce errors in the carbon beam distribution between the treatment planning system and the actual situation, we evaluated the geometrical accuracy, volume accuracy, water-equivalent length (WEL), and treatment planning, and compared the results of evaluation of axial and helical scan methods with various scan parameters. The results indicated that both scan methods showed good geometrical accuracy for thin slice images, but for thick slice images it is easier to understand the phantom as a sphere from the helical as compared with the axial scan. Treatment planning with a thin slice thickness (ST) provided accurate dose distribution for both scan methods, and the dose distribution on the treatment planning system was almost the same as that in the actual situation. Not all institutes, however, can obtain thin slice CT images, and some have used thick slice CT images in planning. For the axial scan, such thick slice images induced differences in dose distribution between treatment planning and the actual situation. Helical scans with a small, reconstructed increment reduced these differences even with relatively thick CT images. To achieve a more accurate dose distribution, radiation therapy planning should be performed using a thin ST for both scan methods or the helical scan with a small, reconstructed increment. Although we reached this conclusion using a carbon beam, it also may be applicable to proton beam therapy.

A novel method of removing artifacts because of metallic dental restorations in 3-D CT images of jaw bone

CT images, especially in a three-dimensional (3-D) mode, give valuable information for oral implant surgery. However, image quality is often severely compromised by artifacts originating from metallic dental restorations, and an effective solution for artifacts is being sought. This study attempts to substitute the damaged areas of the jaw bone images with dental cast model images obtained by CT. The position of the dental cast images was registered to that of the jaw bone images using a devised interface that is composed of an occlusal bite made of self-curing acrylic resin and a marker plate made of gypsum. The patient adapted this interface, and CT images of the stomatognathic system were filmed. On the other hand, this interface was placed between the upper and lower cast models and filmed by CT together with the cast models. The position of the marker plate imaged with the dental casts was registered to those adapted by the patient. The error of registration was examined to be 0.25 mm, which was satisfactory for clinical application. The damaged region in the cranial bone images as an obstacle for implant surgery was removed and substituted with the trimmed images of the dental cast. In the method developed here, the images around the metallic compounds severely damaged by artifacts were successfully reconstructed, and the stomatognathic system images became clear, and this is useful for implant surgery.

An approximate short scan helical FDK cone beam algorithm based on nutating curved surfaces satisfying the Tuy's condition

Traditionally, short scan helical FDK algorithms have been implemented based on horizontal transaxial slices. However, not every point on the horizontal transaxial slice satisfies Tuy's condition for the corresponding (pi+fan angle) segment of helix, which means that some points on the horizontal slices are incompletely sampled and are impossible to be exactly reconstructed. In this paper, we propose and implement an improved but still approximate short scan helical cone beam FDK algorithm based on nutating curved surfaces satisfying the Tuy's condition. This surface is defined by averaging PI surfaces emanating the initial and final source points of a (pi+fan angle) segment of helix. One of the key characteristics of the surface is that every point on it satisfies the Tuy's condition for the corresponding (pi+fan angle) segment of helix, which means that we can potentially reconstruct every point on the surface exactly. This difference makes the proposed algorithm deliver a better-reconstructed image quality while requiring a smaller detector area than that of traditional FDK methods based on horizontal transaxial slices. Another characteristic of the proposed surface is that every point within the helix belongs to one and only one such surface. Therefore, the location of the short scan segment for the reconstruction of a point in Cartesian coordinate could be precalculated and stored in a look-up table. This enables us to perform reconstruction directly on rectangular grids. We compare the performance of the improved FDK algorithm with that of a quasi-exact algorithm based on data combination technique. The simulation results show that the reconstructed image quality of these two methods is similar.

Automatic phase determination for retrospectively gated cardiac CT

The recent improvements in CT detector and gantry technology in combination with new heart rate adaptive cone beam reconstruction algorithms enable the visualization of the heart in three dimensions at high spatial resolution. However, the finite temporal resolution still impedes the artifact-free reconstruction of the heart at any arbitrary phase of the cardiac cycle. Cardiac phases must be found during which the heart is quasistationary to obtain outmost image quality. It is challenging to find these phases due to intercycle and patient-to-patient variability. Electrocardiogram (ECG) information does not always represent the heart motion with an adequate accuracy. In this publication, a simple and efficient image-based technique is introduced which is able to deliver stable cardiac phases in an automatic and patient-specific way. From low-resolution four-dimensional data sets, the most stable phases are derived by calculating the object similarity between subsequent phases in the cardiac cycle. Patient-specific information about the object motion can be determined and resolved spatially. This information is used to perform optimized high-resolution reconstructions at phases of little motion. Results based on a simulation study and three real patient data sets are presented. The projection data were generated using a 16-slice cone beam CT system in low-pitch helical mode with parallel ECG recording.

Multi-detector row CT systems and image-reconstruction techniques

The introduction in 1998 of multi-detector row computed tomography (CT) by the major CT vendors was a milestone with regard to increased scan speed, improved z-axis spatial resolution, and better utilization of the available x-ray power. In this review, the general technical principles of multi-detector row CT are reviewed as they apply to the established four- and eight-section systems, the most recent 16-section scanners, and future generations of multi-detector row CT systems. Clinical examples are used to demonstrate both the potential and the limitations of the different scanner types. When necessary, standard single-section CT is referred to as a common basis and starting point for further developments. Another focus is the increasingly important topic of patient radiation exposure, successful dose management, and strategies for dose reduction. Finally, the evolutionary steps from traditional single-section spiral image-reconstruction algorithms to the most recent approaches toward multisection spiral reconstruction are traced.

Helical cardiac cone beam CT reconstruction with large area detectors: a simulation study

Retrospectively gated cardiac volume CT imaging has become feasible with the introduction of heart rate adaptive cardiac CT reconstruction algorithms. The development in detector technology and the rapid introduction of multi-row detectors has demanded reconstruction schemes which account for the cone geometry. With the extended cardiac reconstruction (ECR) framework, the idea of approximate helical cone beam CT has been extended to be used with retrospective gating, enabling heart rate adaptive cardiac cone beam reconstruction. In this contribution, the ECR technique is evaluated for systems with an increased number of detector rows, which leads to larger cone angles. A simulation study has been carried out based on a 4D cardiac phantom consisting of a thorax model and a dynamic heart insert. Images have been reconstructed for different detector set-ups. Reconstruction assessment functions have been calculated for the detector set-ups employing different rotation times, relative pitches and heart rates. With the increased volume coverage of large area detector systems, low-pitch scans become feasible without resulting in extensive scan times, inhibiting single breath hold acquisitions. ECR delivers promising image results when being applied to systems with larger cone angles.

The frequency split method for helical cone-beam reconstruction

A new approximate method for the utilization of redundant data in helical cone-beam CT is presented. It is based on the observation that the original WEDGE method provides excellent image quality if only little more than 180 degrees data are used for back-projection, and that significant low-frequency artifacts appear if a larger amount of redundant data are used. This degradation is compensated by the frequency split method: The low-frequency part of the image is reconstructed using little more than 180 degrees of data, while the high frequency part is reconstructed using all data. The resulting algorithm shows no cone-beam artifacts in a simulation of a 64-row scanner. It is further shown that the frequency split method hardly degrades the signal-to-noise ratio of the reconstructed images and that it behaves robustly in the presence of motion.

Artifact analysis and reconstruction improvement in helical cardiac cone beam CT

With the introduction of cone beam (CB) scanners, cardiac volumetric computed tomography (CT) imaging has the potential to become a noninvasive imaging tool in clinical routine for the diagnosis of various heart diseases. Heart rate adaptive reconstruction schemes enable the reconstruction of high-resolution volumetric data sets of the heart. Artifacts, caused by strong heart rate variations, high heart rates and obesity, decrease the image quality and the diagnostic value of the images. The image quality suffers from streak artifacts if suboptimal scan and reconstruction parameters are chosen, demanding improved gating techniques. In this paper, an artifact analysis is carried out which addresses the artifacts due to the gating when using a three-dimensional CB cardiac reconstruction technique. An automatic and patient specific cardiac weighting technique is presented in order to improve the image quality. Based on the properties of the reconstruction algorithm, several assessment techniques are introduced which enable the quantitative determination of the cycle-to-cycle transition smoothness and phase homogeneity of the image reconstruction. Projection data of four patients were acquired using a 16-slice CBCT system in low pitch helical mode with parallel electrocardiogram recording. For each patient, image results are presented and discussed in combination with the assessment criteria.

Extended parallel backprojection for standard three-dimensional and phase-correlated four-dimensional axial and spiral cone-beam CT with arbitrary pitch, arbitrary cone-angle, and 100% dose usage

We have developed a new approximate Feldkamp-type algorithm that we call the extended parallel backprojection (EPBP). Its main features are a phase-weighted backprojection and a voxel-by-voxel 180 degrees normalization. The first feature ensures three-dimensional (3-D) and 4-D capabilities with one and the same algorithm; the second ensures 100% detector usage (each ray is accounted for). The algorithm was evaluated using simulated data of a thorax phantom and a cardiac motion phantom for scanners with up to 256 slices. Axial (circle and sequence) and spiral scan trajectories were investigated. The standard reconstructions (EPBPStd) are of high quality, even for as many as 256 slices. The cardiac reconstructions (EPBPCI) are of high quality as well and show no significant deterioration of objects even far off the center of rotation. Since EPBPCI uses the cardio interpolation (CI) phase weighting the temporal resolution is equivalent to that of the well-established single-slice and multislice cardiac approaches 180 degrees CI, 180 degrees MCI, and ASSRCI, respectively, and lies in the order of 50 to 100 ms for rotation times between 0.4 and 0.5 s. EPBP appears to fulfill all required demands. Especially the phase-correlated EPBP reconstruction of cardiac multiple circle scan data is of high interest, e.g., for dynamic perfusion studies of the heart.

On two versions of a 3 algorithm for spiral CT

A 3pi algorithm is obtained in which all the derivatives are confined to a detector array. Distance weighting of backprojection coefficients of the algorithm is studied. A numerical experiment indicates that avoiding differentiation along the source trajectory improves spatial resolution. Another numerical experiment shows that the terms depending on the non-standard distance weighting l/[x - y (s)] can no longer be ignored.

A quasiexact reconstruction algorithm for helical CT using a 3-Pi acquisition

Recently, an exact reconstruction method for helical CT was published by A. Katsevich. The algorithm is of the filtered backprojection type and is, therefore, computationally efficient. Moreover, during backprojection, only data are used which correspond to an illumination interval of 180 degrees as seen from the object-point. We propose a new reconstruction method, which is applicable to data obtained with a 3-Pi acquisition [IEEE Trans. Med. Imaging 19, 848-863 (2000)]. The method uses the same filter types as the Katsevich algorithm, but the directions and the number of the filter lines are chosen differently. For the derivation of the new algorithm, we analyze the relationship of the Katsevich method and radon inversion. A certain radon plane can intersect with the backprojection interval related to a 3-Pi acquisition either once, three, or five times. In analogy to the definition of quasiexactness introduced by Kudo et al. for a 1-Pi acquisition, we use the term quasiexactness for algorithms on a 3-Pi acquisition, if radon planes with one or three intersections within the backprojection interval are treated correctly. Using the results on the relationship with radon inversion, we can prove that our algorithm is quasiexact in this sense. We use simulation results in order to demonstrate that the algorithm yields excellent image quality.

Adaptive temporal resolution optimization in helical cardiac cone beam CT reconstruction

Cone beam computed tomography scanners in combination with heart rate adaptive reconstruction schemes have the potential to enable cardiac volumetric computed tomography (CT) imaging for a larger number of patients and applications. In this publication, an adaptive scheme for the automatic and patient-specific reconstruction optimization is introduced to improve the temporal resolution and image quality. The optimization method permits the automatic determination of the required amount of gated helical cone beam projection data for the reconstruction volume. It furthermore allows one to optimize subvolume reconstruction yielding an increased temporal resolution. In addition, methods for the assessment of the temporal resolution are given which enable a quantitative documentation of the reconstruction improvements. Results are presented for patient data sets acquired in low pitch helical mode using a 16-slice cone beam CT system with parallel ECG recording.

Helical cardiac cone beam reconstruction using retrospective ECG gating

In modern computer tomography (CT) systems, the fast rotating gantry and the increased detector width enable 3D imaging of the heart. Cardiac volume CT has a high potential for non-invasive coronary angiography with high spatial resolution and short scan time. Due to the increased detector width, true cone beam reconstruction methods are needed instead of adapted 2D reconstruction schemes. In this paper, the extended cardiac reconstruction method is introduced. It integrates the idea of retrospectively gated cardiac reconstruction for helical data acquisition into a cone beam reconstruction framework. It leads to an efficient and flexible algorithmic scheme for the reconstruction of single- and multi-phase cardiac volume datasets. The method automatically adapts the number of cardiac cycles used for the reconstruction. The cone beam geometry is fully taken into account during the reconstruction process. Within this paper, results are presented on patient datasets which have been acquired using a 16-slice cone beam CT system.

Image reconstruction and image quality evaluation for a 16-slice CT scanner

We present a theoretical overview and a performance evaluation of a novel approximate reconstruction algorithm for cone-beam spiral CT, the adaptive multiple plane reconstruction (AMPR), which has been introduced by Schaller, Flohr et al. [Proc. SPIE Int. Symp. Med. Imag. 4322, 113-127 (2001)] AMPR has been implemented in a recently introduced 16-slice CT scanner. We present a detailed algorithmic description of AMPR which allows for a free selection of the spiral pitch. We show that dose utilization is better than 90% independent of the pitch. We give an overview on the z-reformation functions chosen to allow for a variable selection of the spiral slice width at arbitrary pitch values. To investigate AMPR image quality we present images of anthropomorphic phantoms and initial patient results. We present measurements of spiral slice sensitivity profiles (SSPs) and measurements of the maximum achievable transverse resolution, both in the isocenter and off-center. We discuss the pitch dependence of image noise measured in a centered 20 cm water phantom. Using the AMPR approach, cone-beam artifacts are considerably reduced for the 16-slice scanner investigated. Image quality in MPRs is independent of the pitch and equivalent to a single-slice CT system at pitch p approximately 1.5. The full width at half-maximum (FWHM) of the spiral SSPs shows only minor variations as a function of the pitch, nominal, and measured values differ by less than 0.2 mm. With 16 x 0.75 mm collimation, the measured FWHM of the smallest reconstructed slice is about 0.9 mm. Using this slice width and overlapping image reconstruction, cylindrical holes with 0.6 mm diameter can be resolved in a z-resolution phantom. Image noise for constant effective mAs is nearly independent of the pitch. Measured and theoretically expected dose utilization are in good agreement. Meanwhile, clinical practice has demonstrated the excellent image quality and the increased diagnostic capability that is obtained with the new generation of multislice CT systems.

Artifacts associated with implementation of the Grangeat formula

To compensate for image artifacts introduced in approximate cone-beam reconstruction, exact cone-beam reconstruction algorithms are being developed for medical x-ray CT. Although the exact cone-beam approach is theoretically error-free, it is subject to image artifacts due to the discrete nature of numerical implementation. We report a study on image artifacts associated with the Grangeat algorithm as applied to a circular scanning locus. Three types of artifacts are found, which are thorn, wrinkle, and V-shaped artifacts. The thorn pattern is created by inappropriate extrapolation into the shadow zone in the radon domain. If the shadow zone is filled in with continuous data, the thorn artifacts along the boundary of the shadow zone can be removed. The wrinkle appearance arises if interpolated first derivatives of the radon data are not smooth between adjacent detector planes. In particular, the nearest-neighbor interpolation method should not be used. If the number of projections is not small, the bilinear interpolation method is effective to suppress the wrinkle artifacts. The V-shaped artifacts on the meridian plane come from the line integrations through the transition zones where derivative data change abruptly. Two remedies are to increase the sampling rate and suppress data noise.