Multi-atlas segmentation of the cartilage in knee MR images with sequential volume- and bone-mask-based registrations

  • Han Sang Lee
  • , Hyeun A. Kim
  • , Hyeonjin Kim
  • , Helen Hong
  • , Young Cheol Yoon
  • , Junmo Kim

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

3 Scopus citations

Abstract

In spite of its clinical importance in diagnosis of osteoarthritis, segmentation of cartilage in knee MRI remains a challenging task due to its shape variability and low contrast with surrounding soft tissues and synovial fluid. In this paper, we propose a multi-atlas segmentation of cartilage in knee MRI with sequential atlas registrations and locallyweighted voting (LWV). First, bone is segmented by sequential volume- and object-based registrations and LWV. Second, to overcome the shape variability of cartilage, cartilage is segmented by bone-mask-based registration and LWV. In experiments, the proposed method improved the bone segmentation by reducing misclassified bone region, and enhanced the cartilage segmentation by preventing cartilage leakage into surrounding similar intensity region, with the help of sequential registrations and LWV.

Original languageEnglish
Title of host publicationMedical Imaging 2016
Subtitle of host publicationComputer-Aided Diagnosis
EditorsGeorgia D. Tourassi, Samuel G. Armato
PublisherSPIE
ISBN (Electronic)9781510600201
DOIs
StatePublished - 2016
EventMedical Imaging 2016: Computer-Aided Diagnosis - San Diego, United States
Duration: 28 Feb 20162 Mar 2016

Publication series

NameProgress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume9785
ISSN (Print)1605-7422

Conference

ConferenceMedical Imaging 2016: Computer-Aided Diagnosis
Country/TerritoryUnited States
CitySan Diego
Period28/02/162/03/16

Keywords

  • Cartilage segmentation
  • Magnetic resonance imaging
  • Multi-atlas segmentation

Fingerprint

Dive into the research topics of 'Multi-atlas segmentation of the cartilage in knee MR images with sequential volume- and bone-mask-based registrations'. Together they form a unique fingerprint.

Cite this