We present a novel method for tracking myocardial motion from volumetric ultrasound data based on non-rigid image registration using an anatomical free-form deformation model. Traditionally, the B-spline control points of such a model are defined on a rectangular grid in Cartesian space. This arrangement may be suboptimal as it treats the blood pool and myocardium similarly and as it enforces spatial smoothness in non-physiological directions. In this work, the basis functions are locally oriented along the radial, longitudinal and circumferential direction of the endocardium. This formulation allows us to model the left ventricular motion more naturally. We obtained encouraging accuracy results for the simulated models, with average errors of 0.8±0.6mm (10% relatively) and 0.5±0.4mm (15% relatively) compared to the ground truth in high and low contractility models respectively. © 2013 Springer-Verlag.
CITATION STYLE
Heyde, B., Barbosa, D., Claus, P., Maes, F., & D’Hooge, J. (2013). Three-dimensional cardiac motion estimation based on non-rigid image registration using a novel transformation model adapted to the heart. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 7746 LNCS, pp. 142–150). https://doi.org/10.1007/978-3-642-36961-2_17
Mendeley helps you to discover research relevant for your work.