Skip to Main content Skip to Navigation
Theses

Topology Preserving Vascular Registration: Application to Percutaneous Coronary Intervention

Abstract : In percutaneous coronary interventions, integrating into the live fluoroscopic image vessel calcifications and occlusion information that are revealed in the pre-operative Computed Tomography Angiography can greatly improve guidance of the clinician. Fusing pre- and intra-operative information into a single space aims at taking advantage of two complementary modalities and requires a step of registration that must provide good alignment and relevant correspondences between them. Most of the existing 3D/2D vessel registration algorithms do not take into account the particular topology of the vasculature to be matched, resulting into pairings that may be topologically inconsistent along the vasculature. A first contribution consisted in a registration framework dedicated to curve matching, denoted the Iterative Closest Curve (ICC). Its main feature is to preserve the topological consistency along curves by taking advantage of the Frechet distance that not only computes the distance between two curves but also builds ordered pairings along them. A second contribution is a pairing procedure designed for the matching of a vascular tree structure that endorses its particular topology and that can easily take advantage of the ICC-framework. Centerlines of the 3D tree are matched to curves extracted from the 2D vascular graph while preserving the connectivity at 3D bifurcations. The matching criterion used to build the pairings takes into account the geometric distance and the resemblance between curves both based on a global formulation using the Frechet distance. To evaluate our approach we run experiments on a database composed of 63 clinical cases, measuring accuracy on real conditions and robustness with respect to a simulated displacement. Quantitative results have been obtained using two complementary measures that aim at assessing the results both geometrically and topologically, and quantify the resulting alignment error as well as the pairing error. The proposed method exhibits good results both in terms of pairing and alignment and demonstrates to be low sensitive to the rotations to be compensated (up to 30 degrees).
Complete list of metadata

Cited literature [70 references]  Display  Hide  Download

https://hal.inria.fr/tel-01235141
Contributor : Thomas Benseghir <>
Submitted on : Saturday, November 28, 2015 - 12:13:07 PM
Last modification on : Thursday, January 17, 2019 - 1:48:04 PM
Long-term archiving on: : Saturday, April 29, 2017 - 12:05:26 AM

Identifiers

  • HAL Id : tel-01235141, version 1

Collections

Citation

Thomas Benseghir. Topology Preserving Vascular Registration: Application to Percutaneous Coronary Intervention. Signal and Image processing. Université de Nice-Sophia Antipolis, 2015. English. ⟨tel-01235141⟩

Share

Metrics

Record views

452

Files downloads

622