In this paper, a novel 2D-3D cardiac image registration algorithm is proposed for application in X-ray-guided catheterisation procedures, and relies on a common technique of inserting a catheter and then looping it inside a chamber of the heart for visual reference. Registration starts with the isocentre-supine constraint and then iteratively refined by maximising a feature-based area metric using an inserted catheter loop and the segmented cardiac border from one or more X-ray views. Maximisation is done in two stages: first correcting for translational motion, and then simultaneously correcting for rotations and translations. The two-staged approach is demonstrated to be more accurate than a similar single-staged approach in an explanted porcine heart. In this experiment, accuracy was demonstrated to be within the 5-mm clinical requirement. On average, the algorithm could register images with a mean target registration error (TRE) of 4.6-mm when using two X-rays (biplane), and a mean reprojection distance (RPD) of 1.9 mm using a single view (monoplane).
CITATION STYLE
Truong, M. V. N., Liu, A., Housden, R. J., Penney, G. P., Pop, M., & Rhode, K. S. (2017). Novel looped-catheter-based 2D-3D registration algorithm for MR, 3DRX and X-ray images: Validation study in an ex-vivo heart. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 10124 LNCS, pp. 152–162). Springer Verlag. https://doi.org/10.1007/978-3-319-52718-5_17
Mendeley helps you to discover research relevant for your work.