Conference Proceeding
Authors: Jing Wu (Medical University of Vienna) , Bianca S. Gerendas (Medical University of Vienna) , Sebastian M. Waldstein (Medical University of Vienna) , Georg Langs (Medical University of Vienna) , Christian Simader (Medical University of Vienna) , Ursula Schmidt-Erfurth (Medical University of Vienna)
We propose a multiple scanner vendor registration method for pathological retinal 3D spectral domain optical coherence tomography volumes based on Myronenko’s Coherent Point Drift and our automated vessel shadow segmentation. Coherent point drift is applied to the segmented retinal vessel point sets used as landmarks to generate the registration parameters required. In contrast to other registration methods, our solution incorporates a landmark detection and extraction method that specifically limits the extraction of false positives and a registration method capable of handling any such noise in the landmark point sets. Our experiments show modified Hausdorff distance is reduced by a minimum of 91% between target and registered vessel point sets with at least 94% of bifurcations correctly overlapping based on ground truth, a significant improvement over current methods.
Keywords:
How to Cite: Wu, J. , Gerendas, B. S. , Waldstein, S. M. , Langs, G. , Simader, C. & Schmidt-Erfurth, U. (2014) “Stable registration of pathological 3D-OCT scans using retinal vessels”, Proceedings of the Ophthalmic Medical Image Analysis International Workshop. 1(2014). doi: https://doi.org/10.17077/omia.1002