Title: Sparse Group Composition for Robust Left Ventricular
Author: Bing Wang Xiaomeng Gu Chonghao Fan Hongzhi
Xie Shuyang Zhang Xuedong Tian Lixu Gu
Reference: CMIG 1384
To appear in: Computerized Medical Imaging and Graphics
Received date: 23-1-2015
Revised date: 20-6-2015
Accepted date: 22-6-2015
Please cite this article as: Wang B, Gu X, Fan C, Xie H, Zhang S,
Tian X, Gu L, Sparse Group Composition for Robust Left Ventricular
Epicardium Segmentation, Computerized Medical Imaging and Graphics (2015), http://dx.doi.org/10.1016/j.compmedimag.2015.06.003
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Page 1 of 14
Ac ce pte d M an us cri pt
Sparse Group Composition for Robust Left Ventricular
Bing Wanga, Xiaomeng Gub, Chonghao Fanb, Hongzhi Xiec*, Shuyang Zhangc,
Xuedong Tiand, Lixu Gub,e* aCollege of Mathematics and Information Science, Hebei University, China bMulti-disciplinary Research Center, Hebei University, China cDepartment of Cardiovascular, Peking Union Medical College Hospital, China dCollege of Computer Science, Hebei University, China eSchool of Biomedical Engineering, Shanghai Jiao Tong University, China
We propose a sparse group composition model (SGC) to model multi-shape prior.
Chan-Vese model with SGC-based refinement is used to segment LV epicardium.
Epi- and endocardium are modeled together removing complex landmark detection.
Epi- and endocardium are modeled together to enhance robustness of segmentation.
Left ventricular (LV) epicardium segmentation in cardiac magnetic resonance images (MRIs) is still a challenging task, where the a-priori knowledge like those that incorporate the heart shape model is usually used to derive reasonable segmentation results. In this paper, we propose a sparse group composition (SGC) approach to model multiple shapes simultaneously, which extends conventional sparsity-based single shape prior modeling to incorporate a-priori spatial constraint information among multiple shapes on-the-fly. Multiple interrelated shapes (shapes of epi- and endo-cardium of myocardium in the case of LV epicardium segmentation) are regarded as a group, and sparse linear composition of training groups is computed to approximate the input group. A framework of iterative procedure of refinement based on
SGC and segmentation based on deformation model is utilized for LV epicardium segmentation, in which an improved shape-constraint gradient Chan-Vese model (GCV) acted as deformation model. Compared with the standard sparsity-based single shape prior modeling, the refinement procedure has strong robust for relative gross and not much sparse errors in the input shape and the initial epicardium location can be estimated without complicated landmark detection due to modeling spatial constraint information among multiple shapes effectively. Proposed method was validated on 45 cardiac cine-MR clinical datasets and the results were compared with expert contours. The average perpendicular distance (APD) error of contours is 1.50±0.29 mm, and the dice metric (DM) is 0.96±0.01. Compared to the
E-mail address: email@example.com, firstname.lastname@example.org.
Page 2 of 14
Ac ce pte d M an us cri pt state-of-the-art methods, our proposed approach appealed competitive segmentation performance and improved robustness.
Keywords: LV epicardium segmentation; sparse group composition; multi-shape prior modeling; gradient Chan-Vese (GCV) model 1. Introduction
Cardiovascular diseases (CVDs) are the number one cause of death globally . An estimated 17.3 million people died from CVDs in 2008, representing 30% of all global deaths . By 2030, the number of people who die from CVDs will increase to reach 23.3 million [1, 2]. Non-invasive assessment of left ventricular function based on cardiac MRI is of great value for the diagnosis and treatment monitoring of these pathologies. For example, LV contractile function quantified through ejection fraction, myocardium mass and ventricle volume, are often used as a crucial indicator in the assessment of deficient blood supply to the cardiac tissue [3, 4].Calculations of such measurements is dependent on accurate delineation of LV myocardial boundaries. However, reliable and accurate automatic delineation of cardiac inner wall and outer wall remains a difficult problem, due to intensity inhomogeneities of tissues outside myocardium, the poor contrast between these tissues and myocardium, papillary muscles connected to the inner myocardium wall, artifacts arising from flow and noise . Some unreliable appearance cues in cardiac
MRI images are demonstrated in Fig.1.
Fig.1. Demonstration of several unreliable appearance cues in cardiac MRI images. (a) Missing border is pointed out by red arrows.
Fuzzy borders are pointed out by white arrows. (b) Papillary muscles within the LV cavity are indicated by green arrows. (c) Muscles connected to the inner myocardium wall are surrounded by yellow ellipses.
Significant numbers of methods have been proposed for (semi-)automated LV segmentation, including approaches using no, weak, strong prior knowledge . Methods that work with weak or no prior knowledge, including methods based on thresholding , dynamic programming [6, 7, 8, 9], clustering [10, 11, 12], and a combination of these [13, 14, 15], have been considered as general LV segmentation methods. Besides, graph -based methods, such as graph cut [16, 17, 18] and random walk , have been introduced that can achieve desired results with limited user-interaction. And, deformable model [20, 21], such as active contour (or snake) model [22, 23], level set , and their variants, have been prevalently applied in LV segmentation. However, almost all of these methods require manual intervention more or less.