Limits...
GPU-based 3D cone-beam CT image reconstruction for large data volume.

Zhao X, Hu JJ, Zhang P - Int J Biomed Imaging (2009)

Bottom Line: For extra large data volume exceeding the physical graphic memory of GPU, a straightforward compromise is to divide data volume into blocks.Different from the conventional Octree partition method, a new partition scheme is proposed in this paper.The method is evaluated by reconstructing 3D images from computer-simulation data and real micro-CT data.

View Article: PubMed Central - PubMed

Affiliation: School of Mathematical Sciences, Capital Normal University, Beijing 100048, China. zhaoxing 1999@yahoo.com

ABSTRACT
Currently, 3D cone-beam CT image reconstruction speed is still a severe limitation for clinical application. The computational power of modern graphics processing units (GPUs) has been harnessed to provide impressive acceleration of 3D volume image reconstruction. For extra large data volume exceeding the physical graphic memory of GPU, a straightforward compromise is to divide data volume into blocks. Different from the conventional Octree partition method, a new partition scheme is proposed in this paper. This method divides both projection data and reconstructed image volume into subsets according to geometric symmetries in circular cone-beam projection layout, and a fast reconstruction for large data volume can be implemented by packing the subsets of projection data into the RGBA channels of GPU, performing the reconstruction chunk by chunk and combining the individual results in the end. The method is evaluated by reconstructing 3D images from computer-simulation data and real micro-CT data. Our results indicate that the GPU implementation can maintain original precision and speed up the reconstruction process by 110-120 times for circular cone-beam scan, as compared to traditional CPU implementation.

No MeSH data available.


Axis-aligned stack of 2D-textured slices for representing reconstructed volume. (a) along Xv axis aligned stack,  (b) along Zv   axis aligned stack, and (c) along the rotation axis Yv aligned stack.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2734932&req=5

fig1: Axis-aligned stack of 2D-textured slices for representing reconstructed volume. (a) along Xv axis aligned stack, (b) along Zv axis aligned stack, and (c) along the rotation axis Yv aligned stack.

Mentions: Current GPUs can be used either as a graphical pipeline or as a multiprocessor chip thanks to the CUDA interface from Nvidia. For both options, the acceleration factor of GPU is high. Xu and Mueller have observed that an implementation of the cone beam back-projection using the graphics pipeline is 3 times faster than the one made with CUDA interface [12]. Hence we use the graphics pipeline to accelerate CT reconstruction in this paper. In order to harness GPU to provide acceleration of 3D volume image reconstruction, we represent reconstructed volume as an axis-aligned stack of 2D-textured slices. The volume may be represented by three kinds of proxy geometries as shown in Figure 1. If the stack of 2D-textured slices aligned along the rotation axis Yv (Figure 1(c)) is adopted, only one data set is enough for circular cone-beam reconstruction. Otherwise, two copies of the data set should be used simultaneously in GPU memory for decreasing the inconsistent sampling rate of volume. This can cause bottlenecks when the memory bandwidth is less than the compute bandwidth, and also needs to merge the two textured slices stacks in each backward-projection loop [13]. Hence, we choose the model shown in Figure 1(c) as reconstructed volume model.


GPU-based 3D cone-beam CT image reconstruction for large data volume.

Zhao X, Hu JJ, Zhang P - Int J Biomed Imaging (2009)

Axis-aligned stack of 2D-textured slices for representing reconstructed volume. (a) along Xv axis aligned stack,  (b) along Zv   axis aligned stack, and (c) along the rotation axis Yv aligned stack.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2734932&req=5

fig1: Axis-aligned stack of 2D-textured slices for representing reconstructed volume. (a) along Xv axis aligned stack, (b) along Zv axis aligned stack, and (c) along the rotation axis Yv aligned stack.
Mentions: Current GPUs can be used either as a graphical pipeline or as a multiprocessor chip thanks to the CUDA interface from Nvidia. For both options, the acceleration factor of GPU is high. Xu and Mueller have observed that an implementation of the cone beam back-projection using the graphics pipeline is 3 times faster than the one made with CUDA interface [12]. Hence we use the graphics pipeline to accelerate CT reconstruction in this paper. In order to harness GPU to provide acceleration of 3D volume image reconstruction, we represent reconstructed volume as an axis-aligned stack of 2D-textured slices. The volume may be represented by three kinds of proxy geometries as shown in Figure 1. If the stack of 2D-textured slices aligned along the rotation axis Yv (Figure 1(c)) is adopted, only one data set is enough for circular cone-beam reconstruction. Otherwise, two copies of the data set should be used simultaneously in GPU memory for decreasing the inconsistent sampling rate of volume. This can cause bottlenecks when the memory bandwidth is less than the compute bandwidth, and also needs to merge the two textured slices stacks in each backward-projection loop [13]. Hence, we choose the model shown in Figure 1(c) as reconstructed volume model.

Bottom Line: For extra large data volume exceeding the physical graphic memory of GPU, a straightforward compromise is to divide data volume into blocks.Different from the conventional Octree partition method, a new partition scheme is proposed in this paper.The method is evaluated by reconstructing 3D images from computer-simulation data and real micro-CT data.

View Article: PubMed Central - PubMed

Affiliation: School of Mathematical Sciences, Capital Normal University, Beijing 100048, China. zhaoxing 1999@yahoo.com

ABSTRACT
Currently, 3D cone-beam CT image reconstruction speed is still a severe limitation for clinical application. The computational power of modern graphics processing units (GPUs) has been harnessed to provide impressive acceleration of 3D volume image reconstruction. For extra large data volume exceeding the physical graphic memory of GPU, a straightforward compromise is to divide data volume into blocks. Different from the conventional Octree partition method, a new partition scheme is proposed in this paper. This method divides both projection data and reconstructed image volume into subsets according to geometric symmetries in circular cone-beam projection layout, and a fast reconstruction for large data volume can be implemented by packing the subsets of projection data into the RGBA channels of GPU, performing the reconstruction chunk by chunk and combining the individual results in the end. The method is evaluated by reconstructing 3D images from computer-simulation data and real micro-CT data. Our results indicate that the GPU implementation can maintain original precision and speed up the reconstruction process by 110-120 times for circular cone-beam scan, as compared to traditional CPU implementation.

No MeSH data available.