Segmentation of Left Ventricle in Cardiac MRI via Contrast-Invariant Deformable Template
- Authors
- Koo, Ja-Keoung; Sohn, Bong Soo; Hong, Byung-Woo
- Issue Date
- Dec-2017
- Publisher
- AMER SCIENTIFIC PUBLISHERS
- Keywords
- Left Ventricle Segmentation; Cardiac MRI; Deformable Template Matching
- Citation
- JOURNAL OF MEDICAL IMAGING AND HEALTH INFORMATICS, v.7, no.8, pp 1682 - 1688
- Pages
- 7
- Journal Title
- JOURNAL OF MEDICAL IMAGING AND HEALTH INFORMATICS
- Volume
- 7
- Number
- 8
- Start Page
- 1682
- End Page
- 1688
- URI
- https://scholarworks.bwise.kr/cau/handle/2019.sw.cau/3579
- DOI
- 10.1166/jmihi.2017.2275
- ISSN
- 2156-7018
2156-7026
- Abstract
- In this paper, we present an image segmentation algorithm for the delineation of the left ventricle boundary in cardiac MRI images. The difficulty in the detection of the left ventricle in cardiac MRI sequences stems from a large variability of its shape and appearance, which change in time and space. We propose a variational approach that is simple yet effective to deal with complex cardiac motion and intensity changes based on a deformable template framework. The segmentation is obtained by optimizing a transformation from a template to its approximate for the region of interest. The energy functional is designed to consider an invariant property with respect to the dynamic contrast change by alternative approximation of motion and appearance. In characterizing the region of interest based on a deformed template we propose two-phase neighborhood where an immediate local neighboring support effectively improve the characteristic power while a global neighborhood models overall background. The cardiac motion within the region of interest is modeled with a simplified affine motion that demonstrates robustness with respect to irregular complex motions. The expectation-maximization algorithm is used to alternatively obtain transformation of template and its approximate to an image. The robustness and effectiveness of the proposed algorithm is demonstrated using MICCAI dataset and the quantitative evaluation is provided in terms of different metrics including F-measure, Dice metric, average perpendicular distance and Hausdorff distance.
- Files in This Item
- There are no files associated with this item.
- Appears in
Collections - College of Software > Department of Artificial Intelligence > 1. Journal Articles
- College of Software > School of Computer Science and Engineering > 1. Journal Articles
![qrcode](https://api.qrserver.com/v1/create-qr-code/?size=55x55&data=https://scholarworks.bwise.kr/cau/handle/2019.sw.cau/3579)
Items in ScholarWorks are protected by copyright, with all rights reserved, unless otherwise indicated.