GPU based iterative cone-beam CT reconstruction using empty space skipping technique

Zhao, Xing; Zhang, Qikun; Yang, Tao

doi:10.3233/xst-130366

Cited by 19 publications

(9 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the implementations of these iterative methods, the line integrals of images

f^{true(n true)}

and

g^{true(n true)}

were calculated by a ray casting method, and back‐projection in formula (11) was performed by a pixel driven method. The ray‐driven and voxel‐driven techniques favor the current GPU architecture, and can be implemented in a great speed by exploiting the parallelism of GPU (Zhu et al ., , Zhao et al ., ). To reconstruct the above images of size 512 × 512 pixels using 720 projections, each of 512 detector cells, MDIR, E‐ART and E‐SART spent almost the same amount of time (0.165 s) in calculating low‐ and high‐energy polychromatic forward projections and updating current density images.…”

Section: Resultsmentioning

confidence: 97%

An extended simultaneous algebraic reconstruction technique (E‐SART) for X‐ray dual spectral computed tomography

2016

Self Cite

View full text Add to dashboard Cite

X-ray dual spectral computed tomography (DSCT) scans the measured object with two different x-ray energy spectra and the collected polychromatic projections from this procedure can be used to perform basis material decomposition of the object. An iterative method E-ART was recently proposed to produce highly quantitative basis material images for DSCT, but it has the drawback of slow convergence and huge computational costs. Inspired by the E-ART method, this paper proposes an extended simultaneous algebraic reconstruction technique (E-SART) for the basis material decomposition of DSCT and an accelerating strategy for improve the convergence speed of the iterative reconstruction. Compared with the E-ART method, the proposed method has much faster convergence rate while at the same time it has better noise suppressing feature. The advantages of this method were verified by experiments in which the FORBILD thorax phantom was iteratively reconstructed from noise-free and noisy polychromatic projections. SCANNING 38:599-611, 2016. © 2016 Wiley Periodicals, Inc.

show abstract

“…In the implementations of these iterative methods, the line integrals of images

f^{true(n true)}

and

g^{true(n true)}

Section: Resultsmentioning

confidence: 97%

An extended simultaneous algebraic reconstruction technique (E‐SART) for X‐ray dual spectral computed tomography

2016

Self Cite

View full text Add to dashboard Cite

show abstract

“…All iterative methods involve at least a single back-and forwardprojection, in addition to correcting noise penalties in the reconstruction domain. Recent studies have focused on achieving faster convergence on the convex optimizer; however, at least 20 iterations are required in order to achieve clinically usable image quality (15)(16)(17). Although a significant amount of computational time can be reduced by parallelizing the process of forwardand back-projection operation using a Graphical Processing Unit (GPU), significant (>80%) amount of the time is still spent on calculating forward-and back-projection and the time is significant (>1 s).…”

Section: Discussionmentioning

confidence: 99%

Analytical Low-Dose CBCT Reconstruction Using Non-local Total Variation Regularization for Image Guided Radiation Therapy

Sohn

Kim

et al. 2020

Front. Oncol.

View full text Add to dashboard Cite

Purpose: Conventional iterative low-dose CBCT reconstruction techniques are slow and tend to over-smooth edges through uniform weighting of the image penalty gradient. In this study, we present a non-iterative analytical low-dose CBCT reconstruction technique by restoring the noisy low-dose CBCT projection with the non-local total variation (NLTV) method.Methods: We modeled the low-dose CBCT reconstruction as recovering high quality, high-dose CBCT x-ray projections (100 kVp, 1.6 mAs) from low-dose, noisy CBCT x-ray projections (100 kVp, 0.1 mAs). The restoration of CBCT projections was performed using the NLTV regularization method. In NLTV, the x-ray image is optimized by minimizing an energy function that penalizes gray-level difference between pair of pixels between noisy x-ray projection and denoising x-ray projection. After the noisy projection is restored by NLTV regularization, the standard FDK method was applied to generate the final reconstruction output.Results: Significant noise reduction was achieved comparing to original, noisy inputs while maintaining the image quality comparable to the high-dose CBCT projections. The experimental validations show the proposed NLTV algorithm can robustly restore the noise level of x-ray projection images while significantly improving the overall image quality. The improvement in normalized mean square error (NMSE) and peak signal-to-noise ratio (PSNR) measured from the non-local total variation-gradient projection (NLTV-GPSR) algorithm is noticeable compared to that of uncorrected low-dose CBCT images. Moreover, the difference of CNRs from the gains from the proposed algorithm is noticeable and comparable to high-dose CBCT. Conclusion:The proposed method successfully restores noise degraded, low-dose CBCT projections to high-dose projection quality. Such an outcome is a considerable improvement to the reconstruction result compared to the FDK-based method. In addition, a significant reduction in reconstruction time makes the proposed algorithm more attractive. This demonstrates the potential use of the proposed algorithm for clinical practice in radiotherapy.

show abstract

“…As described in Section 2.1, storing the entire system matrix in a single GPU is infeasible. Existing proposals use dynamic approaches that calculate the system matrix on the y [24,39]. However, these approaches introduce large computational overhead due to repeated on-the-y computations during iterative image reconstruction [11].…”

Section: Sparse Matrix Compression and Spmvmentioning

confidence: 99%

“…For CT image reconstruction, many researchers have also looked into how to leverage GPUs. For example, Pang et al [24] propose a ray-voxel hybrid driven method to map SART onto GPU architecture, while Zhao et al [39] propose a CUDA-based GPU approach that includes empty-space skipping and a multi-resolution technique to accelerate ART. However, both approaches require a signicant amount of memory due to their lack of awareness of the sparse algebra computational paern and the need to have to calculate weighting factors on the y.…”

Section: Gpu-based Ct Image Reconstructionmentioning

confidence: 99%

“…Some try to accelerate ART by modifying the algorithm [10,21,38], while others try to map ART to high-performance computing (HPC) platforms, e.g., multicore processors [15], Beowulf clusters [19], and FPGAs [9]. Recently, GPUs have been employed as viable accelerators for ART-based CT image reconstruction [11,16,24,39], due to their superior energy eciency and performance, e.g., kernelonly speedups of 10x to 15x over CPU-based solutions, when the compressed sparse row (CSR) format is used to compress data and advanced linear algebra subroutines (e.g., cuSPARSE [16]) are used to accelerate computations. However, a practical CT system matrix stored in the CSR format may still exceed a GPU's memory capacity.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An Enhanced Image Reconstruction Tool for Computed Tomography on GPUs

Wang

Feng

et al. 2017

Proceedings of the Computing Frontiers Conference

View full text Add to dashboard Cite

e algebraic reconstruction technique (ART) is an iterative algorithm for CT (i.e., computed tomography) image reconstruction that delivers beer image quality with less radiation dosage than the industry-standard ltered back projection (FBP). However, the high computational cost of ART requires researchers to turn to highperformance computing to accelerate the algorithm. Alas, existing approaches for ART suer from inecient design of compressed data structures and computational kernels on GPUs. us, this paper presents our enhanced CUDA-based CT image reconstruction tool based on the algebraic reconstruction technique (ART) or cuART. It delivers a compression and parallelization solution for ART-based image reconstruction on GPUs. We address the under-performing, but popular, GPU libraries, e.g., cuSPARSE, BRC, and CSR5, on the ART algorithm and propose a symmetrybased CSR format (SCSR) to further compress the CSR data structure and optimize data access for both SpMV and SpMV T via a column-indices permutation. We also propose sorting-based and sorting-free blocking techniques to optimize the kernel computation by leveraging the sparsity paerns of the system matrix. e end result is that cuART can reduce the memory footprint signicantly and enable practical CT datasets to t into a single GPU. e experimental results on a NVIDIA Tesla K80 GPU illustrate that our approach can achieve up to 6.8x, 7.2x, and 5.4x speedups over counterparts that use cuSPARSE, BRC, and CSR5, respectively.

show abstract

GPU based iterative cone-beam CT reconstruction using empty space skipping technique

Cited by 19 publications

References 19 publications

An extended simultaneous algebraic reconstruction technique (E‐SART) for X‐ray dual spectral computed tomography

An extended simultaneous algebraic reconstruction technique (E‐SART) for X‐ray dual spectral computed tomography

Analytical Low-Dose CBCT Reconstruction Using Non-local Total Variation Regularization for Image Guided Radiation Therapy

An Enhanced Image Reconstruction Tool for Computed Tomography on GPUs

Contact Info

Product

Resources

About