Efficient low-dose CT denoising by locally-consistent non-local means (LC-NLM)

Michael Green, Edith M. Marom, Nahum Kiryati, Eli Konen, Arnaldo Mayer

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

The never-ending quest for lower radiation exposure is a major challenge to the image quality of advanced CT scans. Post-processing algorithms have been recently proposed to improve low-dose CT denoising after image reconstruction. In this work,a novel algorithm,termed the locally-consistent non-local means (LC-NLM),is proposed for this challenging task. By using a database of high-SNR CT patches to filter noisy pixels while locally enforcing spatial consistency,the proposed algorithm achieves both powerful denoising and preservation of fine image details. The LC-NLM is compared both quantitatively and qualitatively,for synthetic and real noise,to state-of-the-art published algorithms. The highest structural similarity index (SSIM) were achieved by LC-NLM in 8 out of 10 denoised chest CT volumes. Also,the visual appearance of the denoised images was clearly better for the proposed algorithm. The favorable comparison results,together with the computational efficiency of LC-NLM makes it a promising tool for low-dose CT denoising.

Original languageEnglish
Title of host publicationMedical Image Computing and Computer-Assisted Intervention - MICCAI 2016 - 19th International Conference, Proceedings
EditorsLeo Joskowicz, Mert R. Sabuncu, William Wells, Gozde Unal, Sebastian Ourselin
PublisherSpringer Verlag
Pages423-431
Number of pages9
ISBN (Print)9783319467252
DOIs
StatePublished - 2016

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9902 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Fingerprint

Dive into the research topics of 'Efficient low-dose CT denoising by locally-consistent non-local means (LC-NLM)'. Together they form a unique fingerprint.

Cite this