A Strategic Approach for Cardiac MR Left Ventricle Segmentation
- 44 Downloads
Quantitative evaluation of cardiac function from cardiac magnetic resonance (CMR) images requires the identification of the myocardial walls. This generally requires the clinician to view the image and interactively trace the contours. Especially, detection of myocardial walls of left ventricle is a difficult task in CMR images that are obtained from subjects having serious diseases. An approach to automated outlining the left ventricular contour is proposed. In order to segment the left ventricle, in this paper, a combination of two approaches is suggested. Difference of Gaussian weighting function (DoG) is newly introduced in random walk approach for blood pool (inner contour) extraction. The myocardial wall (outer contour) is segmented out by a modified active contour method that takes blood pool boundary as the initial contour. Promising experimental results in CMR images demonstrate the potentials of our approach.
KeywordsCardiac magnetic resonance image Gaussian weighting function Difference of Gaussian weighting function Random walk Active contour model
We thank Alexander Andreopoulos (Andreopoulos and Tsotsos 2008) for numerous discussions concerning the data-set, and for providing the data set from Department of Diagnostic Imaging of the Hospital for Sick Children in Toronto, Canada.
- Courant R, Hilbert D. Methods of mathematical physics. vol 2. New York: Wiley; 1989.Google Scholar
- Satpathy A, Eng H, Jiang X. Difference of Gaussian edge-texture based background modeling for dynamic traffic conditions. In: LNCS, number 5358. ISVC 2008. p. 406–17.Google Scholar
- Smolka B, Szczepanski M, lataniotis K, Venetsanopoulos A. Random walk approach to noise reduction in images. In: LNCS, number 2124. New York: Springer; 2002. p. 527–36.Google Scholar
- Zhu X, Lafferty J, Ghahramani Z. Combining active learning and semi-supervised learning using gaussian fields and harmonic functions. In: ICML 2003 workshop on the continuum from labeled to unlabeled data in machine learning and data mining. 2003. p. 58–65.Google Scholar