Abstract:

BACKGROUND: Analysis of the left ventricle motion and deformation is based on accurate segmentation of the left ventricle boundaries. However, affected by the strong gradient from tagged lines in tagged cardiac MR images, extracting endocardium boundaries makes itself very difficult.
OBJECTIVE: A new texture analysis method is proposed, based on a minimum-variance energy map, to suppress the effect of tag lines on image segmentation.
METHODS: Firstly, weighted sums of the local minimum and variance were computed, generating a minimum-variance map; Secondly, some block-like artifacts in the map were filtered off, as well as the boundaries were preserved by using median filter; Finally, GVF-snake model was applied to detect the endocardial contour.
RESULTS AND CONCLUSION: According to the distribution characteristics of tag lines in tagged cardiac MR images, a new texture analysis method based on a minimum-variance energy map is proposed, which can effectively remove tag lines. Experimental results proved that the application of GVF-snake model in the energy map generated by the proposed method can well extract the endocardium boundary of the left ventricle in tagged cardiac MR image.