Optimized Parallelization for Nonlocal Means Based Low Dose CT Image Processing.

Libo Zhang,Zhikun Zhuang,Huazhong Shu,Yining Hu,Benqiang Yang,Limin Luo,Yang Chen

doi:10.1155/2015/790313

Abstract

Low dose CT (LDCT) images are often significantly degraded by severely increased mottled noise/artifacts, which can lead to lowered diagnostic accuracy in clinic. The nonlocal means (NLM) filtering can effectively remove mottled noise/artifacts by utilizing large-scale patch similarity information in LDCT images. But the NLM filtering application in LDCT imaging also requires high computation cost because intensive patch similarity calculation within a large searching window is often required to be used to include enough structure-similarity information for noise/artifact suppression. To improve its clinical feasibility, in this study we further optimize the parallelization of NLM filtering by avoiding the repeated computation with the row-wise intensity calculation and the symmetry weight calculation. The shared memory with fast I/O speed is also used in row-wise intensity calculation for the proposed method. Quantitative experiment demonstrates that significant acceleration can be achieved with respect to the traditional straight pixel-wise parallelization.

Highlights

X-ray Computed Tomography (CT) can reflect human attenuation map in millimeter level, in which rich 3D information of tissues, organs or lesions can be provided for clinical diagnosis
To verify the improvement brought by the proposed acceleration method for the nonlocal means (NLM) filtering, we process the same 512 × 512 low dose CT (LDCT) image in Figure 1(a) using the serial algorithm (CPU based), the conventional parallelization algorithm (GPU based), and the improved parallelization algorithm (GPU based)
In this paper we further optimize the parallelization for NLM filtering in CT image processing

Summary

Introduction

X-ray Computed Tomography (CT) can reflect human attenuation map in millimeter level, in which rich 3D information of tissues, organs or lesions can be provided for clinical diagnosis. Distribution and scale features of noise, artifacts, and normal tissues in CT images need to be jointly considered in designing effective postprocessing algorithms [9, 10] It was pointed in [9,10,11,12,13,14] that the nonlocal means (NLM) filtering, which utilizes the information redundancy property, can effectively suppress noise and artifacts without obviously blurring image details. Since noise and artifacts often distribute with prominent amplitudes in LDCT images, a large searching window is practically required to include more structure information in noise/artifact suppression, which implies a large computation cost This will strongly limit its clinical application considering the large workload in current radiology departments. Experiment results on 2D LDCT images demonstrate that the improved parallelization can significantly shorten computation time and, making itself a potentially applicable processing procedure in LDCT imaging

Nonlocal Means Based Low Dose CT Image Processing

CUDA-Based GPU Acceleration for NLM Algorithm

Experimental Results and Analyses

Discussion and Conclusion