Cardiac MRI Tagline Extraction Based on Diffeomorphic Active Contour Algorithm.

Cardiac tagged Magnetic Resonance Imaging (MRI) is readily available in clinical and preclinical scanners and can be used to quantitatively characterize cardiac motion. Because tag lines fade towards the end of cardiac cycle, accurately extracting their location over the full cardiac cycle is particularly challenging. In this work, we present a geometric flow approach to the segmentation of tag lines in cardiac MRI. This algorithm is based on the diffeomorphic active contour algorithm that evolves straight lines placed on the tagged MR images to match taglines. We evaluated the accuracy of the algorithm using the Hausdorff distance, with errors of $$1.62 \text { } (\pm 0.21 \text { SD})$$ pixel units for horizontal tag lines and $$1.79 \text { } (\pm 0.62 \text { SD})$$ pixel units for vertical ones generated from long axis cardiac images. Our algorithm performed well in the presence of noise or large deformations.