Reconstruction Algorithm For Cone-beam CT Research Articles

A new weighting scheme for arc based circle cone-beam CT reconstruction.

In this paper, we present an arc based fan-beam computed tomography (CT) reconstruction algorithm by applying Katsevich's helical CT image reconstruction formula to 2D fan-beam CT scanning data. Specifically, we propose a new weighting function to deal with the redundant data. Our weighting function ϖ(x_,λ) is an average of two characteristic functions, where each characteristic function indicates whether the projection data of the scanning angle contributes to the intensity of the pixel x_. In fact, for every pixel x_, our method uses the projection data of two scanning angle intervals to reconstruct its intensity, where one interval contains the starting angle and another contains the end angle. Each interval corresponds to a characteristic function. By extending the fan-beam algorithm to the circle cone-beam geometry, we also obtain a new circle cone-beam CT reconstruction algorithm. To verify the effectiveness of our method, the simulated experiments are performed for 2D fan-beam geometry with straight line detectors and 3D circle cone-beam geometry with flat-plan detectors, where the simulated sinograms are generated by the open-source software "ASTRA toolbox." We compare our method with the other existing algorithms. Our experimental results show that our new method yields the lowest root-mean-square-error (RMSE) and the highest structural-similarity (SSIM) for both reconstructed 2D and 3D fan-beam CT images.

Low-dose Cone-Beam Computed Tomography Reconstruction through a fast Three-Dimensional Compressed Sensing Method Based on the Three-Dimensional Pseudo-polar Fourier Transform.

Background:Reconstruction of high quality two dimensional images from fan beam computed tomography (CT) with a limited number of projections is already feasible through Fourier based iterative reconstruction method. However, this article is focused on a more complicated reconstruction of three dimensional (3D) images in a sparse view cone beam computed tomography (CBCT) by utilizing Compressive Sensing (CS) based on 3D pseudo polar Fourier transform (PPFT).Method:In comparison with the prevalent Cartesian grid, PPFT re gridding is potent to remove rebinning and interpolation errors. Furthermore, using PPFT based radon transform as the measurement matrix, reduced the computational complexity.Results:In order to show the computational efficiency of the proposed method, we compare it with an algebraic reconstruction technique and a CS type algorithm. We observed convergence in <20 iterations in our algorithm while others would need at least 50 iterations for reconstructing a qualified phantom image. Furthermore, using a fast composite splitting algorithm solver in each iteration makes it a fast CBCT reconstruction algorithm. The algorithm will minimize a linear combination of three terms corresponding to a least square data fitting, Hessian (HS) Penalty and l1 norm wavelet regularization. We named it PP-based compressed sensing-HS-W. In the reconstruction range of 120 projections around the 360° rotation, the image quality is visually similar to reconstructed images by Feldkamp-Davis-Kress algorithm using 720 projections. This represents a high dose reduction.Conclusion:The main achievements of this work are to reduce the radiation dose without degrading the image quality. Its ability in removing the staircase effect, preserving edges and regions with smooth intensity transition, and producing high-resolution, low-noise reconstruction results in low-dose level are also shown.

Machine log file-based dose verification using novel iterative CBCT reconstruction algorithm in commercial software during volumetric modulated arc therapy for prostate cancer patients.

To evaluate the utility of the use of iterative cone-beam computed tomography (CBCT) for machine log file-based dose verification during volumetric modulated arc therapy (VMAT) for prostate cancer patients. All CBCT acquisition data were used to reconstruct images with the Feldkamp-Davis-Kress algorithm (FDK-CBCT) and the novel iterative algorithm (iCBCT). The Hounsfield unit (HU)-electron density curves for CBCT images were created using the Advanced Electron Density Phantom. The I'mRT and anthropomorphic phantoms were irradiated with VMAT after CBCT registration. Subsequently, fourteen prostate cancer patients received VMAT after CBCT registration. Machine log files and both CBCT images were exported to the PerFRACTION software, and a 3D patient dose was reconstructed. Mean dose for planning target volume (PTV), the bladder, and rectum and the 3D gamma analysis were evaluated. For the phantom studies, the variation of HU values was observed at the central position surrounding the bones in FDK-CBCT. There were almost no changes in the difference of doses at the isocenter between measurement and reconstructed dose for planning CT (pCT), FDK-CBCT, and iCBCT. Mean dose differences of PTV, rectum, and bladder between iCBCT and pCT were approximately 2% lower than those between FDK-CBCT and pCT. For the clinical study, average gamma analysis for 2%/2 mm was 98.22% ± 1.07 and 98.81% ± 1.25% in FDK-CBCT and iCBCT, respectively. A similar machine log file-based dose verification accuracy is obtained for FDK-CBCT and iCBCT during VMAT for prostate cancer patients.

The adaptation and investigation of cone-beam CT reconstruction algorithms for horizontal rotation fixed-gantry scans of rabbits

Fixed-gantry radiation therapy has been proposed as a low-cost alternative to the conventional rotating-gantry radiation therapy, that may help meet the rising global treatment demand. Fixed-gantry systems require gravitational motion compensated reconstruction algorithms to produce cone-beam CT (CBCT) images of sufficient quality for image guidance. The aim of this work was to adapt and investigate five CBCT reconstruction algorithms for fixed-gantry CBCT images. The five algorithms investigated were Feldkamp–Davis–Kress (FDK), prior image constrained compressed sensing (PICCS), gravitational motion compensated FDK (GMCFDK), motion compensated PICCS (MCPICCS) (a novel CBCT reconstruction algorithm) and simultaneous motion estimation and iterative reconstruction (SMEIR). Fixed-gantry and rotating-gantry CBCT scans were acquired of 3 rabbits, with the rotating-gantry scans used as a reference. Projections were sorted into rotation bins, based on the angle of rotation of the rabbit during image acquisition. The algorithms were compared using the structural similarity index measure root mean square error, and reconstruction time. Evaluation of the reconstructed volumes showed that, when compared with the reference rotating-gantry volume, the conventional FDK algorithm did not accurately reconstruct fixed-gantry CBCT scans. Whilst the PICCS reconstruction algorithm reduced some motion artefacts, the motion estimation reconstruction methods (GMCFDK, MCPICCS and SMEIR) were able to greatly reduce the effect of motion artefacts on the reconstructed volumes. This finding was verified quantitatively, with GMCFDK, MCPICCS and SMEIR reconstructions having RMSE 17%–19% lower and SSIM 1% higher than a conventional FDK. However, all motion compensated fixed-gantry CBCT reconstructions had a 56%–61% higher RMSE and 1.5% lower SSIM than FDK reconstructions of conventional rotating-gantry CBCT scans. The results show that motion compensation is required to reduce motion artefacts for fixed-gantry CBCT reconstructions. This paper further demonstrates the feasibility of fixed-gantry CBCT scans, and the ability of CBCT reconstruction algorithms to compensate for motion due to horizontal rotation.

Metal artifact reduction using iterative CBCT reconstruction algorithm for head and neck radiation therapy: A phantom and clinical study

PurposeTo investigate whether a novel iterative cone-beam computed tomography (CBCT) reconstruction algorithm reduces metal artifacts in head and neck patient images. MethodAn anthropomorphic phantom and 35 patients with dental metal prostheses or implants were analyzed. All CBCT images were acquired using a TrueBeam linear accelerator and reconstructed with a Feldkamp–Davis–Kress algorithm-based CBCT (FDK-CBCT) and an iterative CBCT algorithm. The mean Hounsfield unit (HU) and standard deviation values were measured on the tongue near the metal materials and the unaffected region as reference values. The artifact index (AI) was calculated. For objective image analysis, the HU value and AI were compared between FDK-CBCT and iterative CBCT images in phantom and clinical studies. Subjective image analyses of metal artifact scores and soft tissue visualizations were conducted using a five-point scale by two reviewers in the clinical study. ResultsThe HU value and AI showed significant artifact reduction for the iterative CBCT than for the FDK-CBCT images (phantom study: 389.8 vs.−10.3 for HU value, 322.9 vs. 96.2 for AI, FDK-CBCT vs. iterative CBCT, respectively; clinical study: 210.3 vs. 69.0 for HU value, 149.6 vs. 70.7 for AI). The subjective scores in the clinical patient study were improved in the iterative CBCT images (metal artifact score: 1.1 vs. 2.9, FDK-CBCT vs. iterative CBCT, respectively; soft tissue visualization: 1.8 vs. 3.6). ConclusionsThe iterative CBCT reconstruction algorithm substantially reduced metal artifacts caused by dental metal prostheses and improved soft tissue visualization compared to FDK-CBCT in phantom and clinical studies.

Development of Sparse Reconstruction Algorithm of Cone-Beam CT

CT technology has been widely used in various fields such as medical treatment, industry, and materials. Recently years, cone-beam CT (CBCT) used in the medical field is replacing the traditional spiral CT slowly due to its unique advantages. Improving the performance of CBCT and the image reconstruction algorithms could obtain higher quality of images, meanwhile, these also reduce the exposure time of X-ray irradiation. And image reconstruction techniques based on sparse angles have benefits for both. This article briefly introduces the advantages of CBCT and the shortcomings of spiral CT, then the traditional filtered projection algorithm Feldkamp is explained. The development of the CBCT reconstruction algorithms based on incomplete data is analyzed from three aspects: TV model, dictionary learning and compressed sensing sampling. The advantages and disadvantages of these algorithms are analyzed for the development of new algorithms.

Element-resolved Multi-energy Cone Beam CT Reconstruction to Enhance Detectability of Low-concentration Iodine Contrast Agent

Cone beam CT (CBCT) is widely used as an imaging guidance tool in radiation therapy. Yet tumor delineation on CBCT is sometimes challenging, such as in liver cancer, leading to tumor positioning uncertainty. Iodine contrast agent is typically used in liver cancer at CT simulation to enhance detectability. Its application in CBCT is still limited due to poor image quality and excessive contrast agent use in multiple treatment fractions. In this study, we develop a multi-energy element-resolved (MEER) CBCT framework achieved on a standard CBCT platform to enhance detectability of low-concentration Iodine contrast agent. We employ a kVp-switching technique to acquire a multi-energy CBCT scan on standard CBCT platform. We develop an element-resolved CBCT reconstruction algorithm that simultaneously reconstructs x-ray attenuation images at each kVp, a density image, and elemental composition images (H, O, Ca and I). The algorithm solves an optimization problem under the constraint that elemental composition of each voxel is a sparse representation of a dictionary containing compositions of typical human tissues and iodine. To demonstrate feasibility and effectiveness of our method, we perform Monte Carlo simulation studies and experimental validations using a Gammex phantom with inserts containing water and iodine solution (175 mgI/mL) of different concentrations. We consider a kVp switching scan with 600 projections acquired in a full gantry rotation under 80, 100, and 120 kVps. Experiment is realized on the on-board imaging system of a linear accelerator available in our clinic. We analyze iodine detectability as quantified by contrast-to-noise ratio (CNR). We compare results with those of single energy CBCT images reconstructed by the standard filter back projection (FBP) method with 600 projections. Results of simulation and experimental study agree with each other. MEER-CBCT and FBP method achieve similar contrast enhancement. At 4% iodine solution concentration, FBP method achieve ∼250 HU enhancement and CNR of ∼2.5, considered to be the standard CNR for successful tumor visualization. In contrast, MEER-CBCT achieve the same CNR but at the 8 times lower iodine concentration of 0.5%. We have developed a novel MEER-CBCT framework realized on a standard CBCT platform to improve detectability of low-concentration iodine. Simulation and experimental studies demonstrated potential of MEER-CBCT to visualize iodine contrast at a 8 times reduction of iodine concentration.Abstract 208; Table 1Iodine Solution ConcentrationSimulationExperimentMEER-CBCTFBPMEER-CBCTFBPEnhancement (HU)CNREnhancement (HU)CNREnhancement (HU)CNREnhancement (HU)CNR0.1%28.03.430.40.421.91.823.40.30.5%48.95.951.60.755.02.557.90.51%74.48.375.91.079.87.076.80.92%127.313.8129.31.7144.112.4140.31.63%179.016.4179.72.4206.412.3202.62.15%280.923.8281.23.7308.925.4305.63.3 Open table in a new tab

Shading correction assisted iterative cone-beam CT reconstruction

Recent advances in total variation (TV) technology enable accurate CT image reconstruction from highly under-sampled and noisy projection data. The standard iterative reconstruction algorithms, which work well in conventional CT imaging, fail to perform as expected in cone beam CT (CBCT) applications, wherein the non-ideal physics issues, including scatter and beam hardening, are more severe. These physics issues result in large areas of shading artifacts and cause deterioration to the piecewise constant property assumed in reconstructed images. To overcome this obstacle, we incorporate a shading correction scheme into low-dose CBCT reconstruction and propose a clinically acceptable and stable three-dimensional iterative reconstruction method that is referred to as the shading correction assisted iterative reconstruction. In the proposed method, we modify the TV regularization term by adding a shading compensation image to the reconstructed image to compensate for the shading artifacts while leaving the data fidelity term intact. This compensation image is generated empirically, using image segmentation and low-pass filtering, and updated in the iterative process whenever necessary. When the compensation image is determined, the objective function is minimized using the fast iterative shrinkage-thresholding algorithm accelerated on a graphic processing unit. The proposed method is evaluated using CBCT projection data of the Catphan© 600 phantom and two pelvis patients. Compared with the iterative reconstruction without shading correction, the proposed method reduces the overall CT number error from around 200 HU to be around 25 HU and increases the spatial uniformity by a factor of 20 percent, given the same number of sparsely sampled projections. A clinically acceptable and stable iterative reconstruction algorithm for CBCT is proposed in this paper. Differing from the existing algorithms, this algorithm incorporates a shading correction scheme into the low-dose CBCT reconstruction and achieves more stable optimization path and more clinically acceptable reconstructed image. The method proposed by us does not rely on prior information and thus is practically attractive to the applications of low-dose CBCT imaging in the clinic.

Three-dimensional weighting reconstruction algorithm for circular cone-beam CT under large scan angles

Improving imaging quality of cone-beam CT under large cone angle scan has been an important area of CT imaging research. Considering the idea of conjugate rays and making up missing data, we propose a three-dimensional (3D) weighting reconstruction algorithm for cone-beam CT. The 3D weighting function is added in the back-projection process to reduce the axial density drop and improve the accuracy of FDK algorithm. Having a simple structure, the algorithm can be implemented easily without rebinning the native cone-beam data into cone-parallel beam data. Performance of the algorithm is evaluated using two computer simulations and a real industrial component, and the results show that the algorithm achieves better performance in reduction of axial intensity drop artifacts and has a wide range of application.

Feasibility study of cone-beam computed tomography (CBCT) reconstruction algorithm for proton image-guided radiation therapy (IGRT)

The radiation therapy is most important treatment technique in the field of the oncology. With the improved radiation therapy technologies such as proton image-guided radiation therapy (IGRT), the accuracy of the patient alignment system has been emphasized. The purpose of this study was to validate the possibility of cone-beam computed tomography (CBCT) reconstruction algorithm using X-ray and 3-D proton IGRT by acquiring images sequentially with continuous 360° gantry rotation. To confirm the feasibility of CBCT reconstruction algorithm, we acquired the reconstruction images of the proton IGRT for a mouse and ball phantom, respectively. According to the results, we demonstrated the feasibility of reconstruction algorithm for the proton IGRT. In conclusion, this reconstruction algorithm can help the determination of a treatment target and will be used a main patient alignment system for the proton IGRT. In the future work we need to compare the 180 projection view using X-ray imaging system of the proton therapy with other efficient projection views.

SU‐F‐J‐198: A Cross‐Platform Adaptation of An a Priori Scatter Correction Algorithm for Cone‐Beam Projections to Enable Image‐ and Dose‐Guided Proton Therapy

Purpose:Cone‐beam CT (CBCT) imaging may enable image‐ and dose‐guided proton therapy, but is challenged by image artefacts. The aim of this study was to demonstrate the general applicability of a previously developed a priori scatter correction algorithm to allow CBCT‐based proton dose calculations.Methods:The a priori scatter correction algorithm used a plan CT (pCT) and raw cone‐beam projections acquired with the Varian On‐Board Imager. The projections were initially corrected for bow‐tie filtering and beam hardening and subsequently reconstructed using the Feldkamp‐Davis‐Kress algorithm (rawCBCT). The rawCBCTs were intensity normalised before a rigid and deformable registration were applied on the pCTs to the rawCBCTs. The resulting images were forward projected onto the same angles as the raw CB projections. The two projections were subtracted from each other, Gaussian and median filtered, and then subtracted from the raw projections and finally reconstructed to the scatter‐corrected CBCTs. For evaluation, water equivalent path length (WEPL) maps (from anterior to posterior) were calculated on different reconstructions of three data sets (CB projections and pCT) of three parts of an Alderson phantom. Finally, single beam spot scanning proton plans (0–360 deg gantry angle in steps of 5 deg; using PyTRiP) treating a 5 cm central spherical target in the pCT were re‐calculated on scatter‐corrected CBCTs with identical targets.Results:The scatter‐corrected CBCTs resulted in sub‐mm mean WEPL differences relative to the rigid registration of the pCT for all three data sets. These differences were considerably smaller than what was achieved with the regular Varian CBCT reconstruction algorithm (1–9 mm mean WEPL differences). Target coverage in the re‐calculated plans was generally improved using the scatter‐corrected CBCTs compared to the Varian CBCT reconstruction.Conclusion:We have demonstrated the general applicability of a priori CBCT scatter correction, potentially opening for CBCT‐based image/dose‐guided proton therapy, including adaptive strategies.Research agreement with Varian Medical Systems, not connected to the present project.

Priori mask guided image reconstruction (p-MGIR) for ultra-low dose cone-beam computed tomography

Recently, the compressed sensing (CS) based iterative reconstruction method has received attention because of its ability to reconstruct cone beam computed tomography (CBCT) images with good quality using sparsely sampled or noisy projections, thus enabling dose reduction. However, some challenges remain. In particular, there is always a tradeoff between image resolution and noise/streak artifact reduction based on the amount of regularization weighting that is applied uniformly across the CBCT volume. The purpose of this study is to develop a novel low-dose CBCT reconstruction algorithm framework called priori mask guided image reconstruction (p-MGIR) that allows reconstruction of high-quality low-dose CBCT images while preserving the image resolution.In p-MGIR, the unknown CBCT volume was mathematically modeled as a combination of two regions: (1) where anatomical structures are complex, and (2) where intensities are relatively uniform. The priori mask, which is the key concept of the p-MGIR algorithm, was defined as the matrix that distinguishes between the two separate CBCT regions where the resolution needs to be preserved and where streak or noise needs to be suppressed. We then alternately updated each part of image by solving two sub-minimization problems iteratively, where one minimization was focused on preserving the edge information of the first part while the other concentrated on the removal of noise/artifacts from the latter part. To evaluate the performance of the p-MGIR algorithm, a numerical head-and-neck phantom, a Catphan 600 physical phantom, and a clinical head-and-neck cancer case were used for analysis. The results were compared with the standard Feldkamp-Davis-Kress as well as conventional CS-based algorithms.Examination of the p-MGIR algorithm showed that high-quality low-dose CBCT images can be reconstructed without compromising the image resolution. For both phantom and the patient cases, the p-MGIR is able to achieve a clinically-reasonable image with 60 projections. Therefore, a clinically-viable, high-resolution head-and-neck CBCT image can be obtained while cutting the dose by 83%. Moreover, the image quality obtained using p-MGIR is better than the quality obtained using other algorithms.In this work, we propose a novel low-dose CBCT reconstruction algorithm called p-MGIR. It can be potentially used as a CBCT reconstruction algorithm with low dose scan requests

CBCT volumetric coverage extension using a pair of complementary circular scans with complementary kV detector lateral and longitudinal offsets

Onboard cone-beam CT (CBCT) has been widely used in image guided radiation therapy. However, the longitudinal coverage is only 15.5 cm in the pelvis scan mode. As a result, a single CBCT scan cannot cover the planning target volume in the longitudinal direction for over 80% of the patients. The common approach is to use double- or multiple-circular scans and then combine multiple CBCT volumes after reconstruction. However it raises concerns regarding doubled imaging dose at the imaging beam junctions due to beam divergence. In this work, we present a new method, DSCS (Dual Scan with Complementary Shifts), to address the CBCT coverage problem using a pair of complementary circular scans. In DSCS, two circular scans were performed at 39.5 cm apart longitudinally. In the superior scan, the detector panel was offset by 16 cm to the left, 15 cm to the inferior. In the inferior scan, the detector panel was shifted 16 cm to the right and 15 cm to the superior. The effective imaging volume is 39.5 cm longitudinally with a 45 cm lateral field-of-view (FOV). Half beam blocks were used to confine the imaging radiation inside the volume of interest. A new image reconstruction algorithm was developed, based on the Feldkamp–Davis–Kress cone-beam CT reconstruction algorithm, to support the DSCS scanning geometry. Digital phantom simulations were performed to demonstrate the feasibility of DSCS. Physical phantom studies were performed using an anthropomorphic phantom on a commercial onboard CBCT system. With basic scattering corrections, the reconstruction results were acceptable. Other issues, including the discrepancy in couch vertical at different couch longitudinal positions, and the inaccuracy in couch table longitudinal movement, were manually corrected during the reconstruction process. In conclusion, the phantom studies showed that, using DSCS, a 39.5 cm longitudinal coverage with a 45 cm FOV was accomplished. The efficiency of imaging dose usage was near 100%. This proposed method could be potentially useful for image guidance and subsequent treatment plan adaptation.

A new voxellation scheme for iterative CBCT reconstruction with less computational cost: Polar voxel design of an isosceles-triangle shape

We propose a new voxellation scheme for iterative cone-beam computed tomography (CBCT) reconstruction with less computational cost. Here, the voxel configuration is designed with an isosceles-triangle shape by using polar coordinates, exploiting the rotational symmetries inherent to the CBCT acquisition geometry and keeping the uniformity of the voxel size as well. By using polar symmetries, we can reduce the size of the system matrix by a factor corresponding to the number of acquired projections, which speeds up the construction of the system matrix and, thus, allows iterative methods to be applied to practical applications within a reasonable reconstruction time. In this study, we implemented an efficient algorithm to reconstruct the system matrix based upon the proposed voxellation scheme and incorporated it into a built-in iterative CBCT reconstruction algorithm based the gradient-projection Barzilai-Borwein (GPBB) method, and we performed systematic simulation works to investigate the imaging characteristics. Our results indicate that the voxellation scheme we proposed in study seems to be effective in reducing the computational time and the memory burden considerably.

SU‐E‐I‐99: Estimation of Effective Charge Distribution by Dual‐Energy CT Reconstruction

Purpose:Computed Tomography (CT) is a method to produce slice image of specific volume from the scanned x‐ray projection images. The contrast of CT image is correlated with the attenuation coefficients of the x‐ray in the object. The attenuation coefficient is strongly dependent on the x‐ray energy and the effective charge of the material. The purpose of this presentation is to show the effective charge distribution predicted by CT images reconstructed with kilovoltage(kV) and megavoltage(MV) x‐ray energy.Methods:The attenuation coefficients of x‐ray can be characterized by cross section of photoionization and Compton scattering for the specific xray energy. In particular, the photoionization cross section is strongly correlated with the effective charge of the object. Hence we can calculate effective charge by solving the coupled equation between the attenuation coefficient and the theoretical cross section. For this study, we use the megavoltage (MV) and kilovoltage (kV) x‐rays of Elekta Synergy as the dual source x‐ray, and CT image of the Phantom Laboratory CatPhan is reconstructed by the filtered back projection (FBP) and iterative algorithm for cone‐beam CT (CBCT).Results:We report attenuation coefficients of each component of the CatPhan specified by each x‐ray source. Also the effective charge distribution is evaluated by the MV and kV dual x‐ray sources. The predicted effective charges are comparable with the nominal ones.Conclusion:We developed the MV and kV dual‐source CBCT reconstruction to yield the effective charge distribution. For more accuracy, it is critical to remove an effect of the scattering photon in the CBCT reconstruction algorithm. The finding will be fine reference of the effective charge of tissue and lead to the more realistic absorbed‐dose calculation.This work was partly supported by the JSPS Core‐to‐Core Program(No. 23003), and this work was partly supported by JSPS KAKENHI 24234567.

WE-G-18A-04: 3D Dictionary Learning Based Statistical Iterative Reconstruction for Low-Dose Cone Beam CT Imaging

Purpose: To develop a 3D dictionary learning based statistical reconstruction algorithm on graphic processing units (GPU), to improve the quality of low-dose cone beam CT (CBCT) imaging with high efficiency. Methods: A 3D dictionary containing 256 small volumes (atoms) of 3×3×3 voxels was trained from a high quality volume image. During reconstruction, we utilized a Cholesky decomposition based orthogonal matching pursuit algorithm to find a sparse representation on this dictionary basis of each patch in the reconstructed image, in order to regularize the image quality. To accelerate the time-consuming sparse coding in the 3D case, we implemented our algorithm in a parallel fashion by taking advantage of the tremendous computational power of GPU. Evaluations are performed based on a head-neck patient case. FDK reconstruction with full dataset of 364 projections is used as the reference. We compared the proposed 3D dictionary learning based method with a tight frame (TF) based one using a subset data of 121 projections. The image qualities under different resolutions in z-direction, with or without statistical weighting are also studied. Results: Compared to the TF-based CBCT reconstruction, our experiments indicated that 3D dictionary learning based CBCT reconstruction is able to recover finer structures, to remove more streaking artifacts, and is less susceptible to blocky artifacts. It is also observed that statistical reconstruction approach is sensitive to inconsistency between the forward and backward projection operations in parallel computing. Using high a spatial resolution along z direction helps improving the algorithm robustness. Conclusion: 3D dictionary learning based CBCT reconstruction algorithm is able to sense the structural information while suppressing noise, and hence to achieve high quality reconstruction. The GPU realization of the whole algorithm offers a significant efficiency enhancement, making this algorithm more feasible for potential clinical application. A high zresolution is preferred to stabilize statistical iterative reconstruction. This work was supported in part by NIH(1R01CA154747-01), NSFC((No. 61172163), Research Fund for the Doctoral Program of Higher Education of China (No. 20110201110011), China Scholarship Council.

3D dictionary learning based iterative cone beam CT reconstruction

Purpose: This work is to develop a 3D dictionary learning based cone beam CT (CBCT) reconstruction algorithm on graphic processing units (GPU) to improve the quality of sparse-view CBCT reconstruction with high efficiency. Methods : A 3D dictionary containing 256 small volumes (atoms) of 3 × 3 × 3 was trained from a large number of blocks extracted from a high quality volume image. On the basis, we utilized cholesky decomposition based orthogonal matching pursuit algorithm to find the sparse representation of each block. To accelerate the time-consuming sparse coding in the 3D case, we implemented the sparse coding in a parallel fashion by taking advantage of the tremendous computational power of GPU. Conjugate gradient least square algorithm was adopted to minimize the data fidelity term. Evaluations are performed based on a head-neck patient case. FDK reconstruction with full dataset of 364 projections is used as the reference. We compared the proposed 3D dictionary learning based method with tight frame (TF) by performing reconstructions on a subset data of 121 projections. Results: Compared to TF based CBCT reconstruction that shows good overall performance, our experiments indicated that 3D dictionary learning based CBCT reconstruction is able to recover finer structures, remove more streaking artifacts and also induce less blocky artifacts. Conclusion : 3D dictionary learning based CBCT reconstruction algorithm is able to sense the structural information while suppress the noise, and hence to achieve high quality reconstruction under the case of sparse view. The GPU realization of the whole algorithm offers a significant efficiency enhancement, making this algorithm more feasible for potential clinical application. ------------------------------- Cite this article as: Bai T, Yan H, Shi F, Jia X, Lou Y, Xu Q, Jiang S, Mou X. 3D dictionary learning based iterative cone beam CT reconstruction. Int J Cancer Ther Oncol 2014; 2(2):020240. DOI: 10.14319/ijcto.0202.40

Accelerating Popular CBCT Reconstruction Algorithms on GPGPU Using OpenCL

Reconstruction of 3D volumetric data from cone beam CT (CBCT) projections is a computationally demanding task. Open Computing Language (OpenCL) is an open framework for parallel programming on heterogeneous computing platforms with high performance and excellent portability and scalability. The purpose of this study is to employ the OpenCL architecture to accelerate the two most popular CBCT reconstruction algorithms: Feldkamp-Davis-Kress (FDK) and simultaneous algebraic reconstruction technique (SART) on general-purpose graphics processing unit (GPGPU).

TH‐C‐103‐06: Non‐Coplanar Cone Beam CT Reconstruction with Limited Angle of Projections Using Rigid Image Registration Facilitated Prior Image Constrained Compressed Sensing

Purpose: The purpose of this work is to develop a non‐coplanar cone beam CT (CBCT) image reconstruction method from limited angle of projections. Methods: Planning CT or CBCT acquired at neutral position (without couch rotation/translation) serves as the prior images for successive reconstruction. During the course of radiation therapy, subsequent CBCT scans are acquired from limited angle of projections. An iterative reconstruction algorithm, using prior image constrained compressed sensing (PICCS) framework and rigid image registration, is incorporated to reconstruct the patient image under non‐coplanar geometry. The prior reconstructed image is rotated/translated according to the nominal couch rotation/translation to serve as the prior input of the compressed sensing based iterative optimization. First, The PICCS framework can be solved iteratively using the total variation minimization step and the iterative image reconstruction step with simultaneous algebraic reconstruction technique (SART). Then, rigid image registration between the reconstructed image and the rotated/translated prior image is applied to further calibrate the rotation/translation of the prior image. The iterations of PICCS algorithm and rigid image registration continue until the registration results are below the predetermined threshold. The proposed reconstruction algorithm is evaluated with both digital phantom simulations and experimental data. Results: The proposed CBCT reconstruction algorithm significantly improved the image quality from projections with angular range as small as 72 degrees. It can reduce the rigid translational setup errors under from 8 mms to below 1 mm and the rigid rotational setup error from 5 degrees to below 1 degree. Compared with the original PICCS algorithm alone, the rigid registration step helps to improve the convergence of algorithm significantly. Conclusions: The proposed algorithm provides an accurate remedy for solving the problem of non‐coplanar CBCT reconstruction from limited angle of projections by taking advantage of the combination of PICCS framework and rigid image registration.

Fast compressed sensing-based CBCT reconstruction using Barzilai-Borwein formulation for application to on-line IGRT

Compressed sensing theory has enabled an accurate, low-dose cone-beam computed tomography (CBCT) reconstruction using a minimal number of noisy projections. However, the reconstruction time remains a significant challenge for practical implementation in the clinic. In this work, we propose a novel gradient projection algorithm, based on the Gradient-Projection-Barzilai-Borwein formulation (GP-BB), that handles the total variation (TV)-norm regularization-based least squares problem for the CBCT reconstruction in a highly efficient manner, with speed acceptable for routine use in the clinic. CBCT is reconstructed by minimizing an energy function consisting of a data fidelity term and a TV-norm regularization term. Both terms are simultaneously minimized by calculating the gradient projection of the energy function with the step size determined using an approximate Hessian calculation at each iteration, based on the Barzilai-Borwein formulation. To speed up the process, a multiresolution optimization is used. In addition, the entire algorithm was designed to run with a single graphics processing unit (GPU) card. To evaluate the performance, the Shepp-Logan numerical phantom, the CatPhan 600 physical phantom, and a clinically-treated head-and-neck patient were acquired from the TrueBeam™ system (Varian Medical Systems, Palo Alto, CA). For each scan, in total, 364 projections were acquired in a 200° rotation. The imager has 1024 × 768 pixels with 0.388 × 0.388-mm resolution. This was down-sampled to 512 × 384 pixels with 0.776 × 0.776-mm resolution for reconstruction. Evenly spaced angles were subsampled and used for varying the number of projections for the image reconstruction. To assess the performance of our GP-BB algorithm, we have implemented and compared with three compressed sensing-type algorithms, the two of which are popular and published (forward-backward splitting techniques), and the other one with a basic line-search technique. In addition, the conventional Feldkamp-Davis-Kress (FDK) reconstruction of the clinical patient data is compared as well. In comparison with the other compressed sensing-type algorithms, our algorithm showed convergence in ≤30 iterations whereas other published algorithms need at least 50 iterations in order to reconstruct the Shepp-Logan phantom image. With the CatPhan phantom, the GP-BB algorithm achieved a clinically-reasonable image with 40 projections in 12 iterations, in less than 12.6 s. This is at least an order of magnitude faster in reconstruction time compared with the most recent reports utilizing GPU technology given the same input projections. For the head-and-neck clinical scan, clinically-reasonable images were obtained from 120 projections in 34-78 s converging in 12-30 iterations. In this reconstruction range (i.e., 120 projections) the image quality is visually similar to or better than the conventional FDK reconstructed images using 364 projections. This represents a dose reduction of nearly 67% (120∕364 projections) while maintaining a reasonable speed in clinical implementation. In this paper, we proposed a novel, fast, low-dose CBCT reconstruction algorithm using the Barzilai-Borwein step-size calculation. A clinically viable head-and-neck image can be obtained within ∼34-78 s while simultaneously cutting the dose by approximately 67%. This makes our GP-BB algorithm potentially useful in an on-line image-guided radiation therapy (IGRT).