Velocity-aided cardiac segmentation

Abstract

Segmentation of myocardial boundaries, especially the endocardial boundary, in images from magnetic resonance imaging (MRI) often suffers from flow-related signal loss, partial volume effects, and the presence of papillary muscles. To address these problems, a velocity-aided cardiac segmentation method based a modified active contour model with the orientation gradient force has been developed. Unlike other MRI cardiac segmentation methods based on active contour models, the velocity images from phase contrast MRI, together with the magnitude images, were used to derive an additional external force from the orientation gradient. The velocity images were also used to track the initial seed contours throughout the entire cardiac cycle to reduce the propagation of errors in sequential segmentation. The linear correlation coefficients and boundary matching descriptors of segmented boundaries were calculated relative to manually segmented reference boundaries to assess the accuracy of this method. Results of segmentation of the endocardial boundaries were encouraging in both individual frame segmentation and sequential frame segmentation.