Landmark-based registration of OCT image sequences for the follow-up of stent coverage and apposition

Laurent Sarry (1) , Romain Dumas (2) , Florian Dubuisson , Emilie Pery (1) , Nicolas Combaret (3) , Pascal Motreff (1)

1 IP - Institut Pascal
2 CREOP - Centre de Recherches sur l'Entreprise, les Organisations et le Patrimoine
3 CHU Gabriel Montpied [Clermont-Ferrand]

We propose a new landmark-based registration method designed to take into account the specific geometry of OCT endovascular acquisition. B-spline approximation is used to recover probe trajectory and angular offsets between two image sequences from a few landmarks pinpointed out by the cardi-ologist. The well-posedness of the control point estimation is assured by a bending regularization energy automatically weighted by leave-one-out cross-validation. The accuracy of the algorithm is validated on phantom datasets and its clinical interest illustrated on patient datasets by comparing coverage and apposition score maps for registered image volumes.

Mots clés: optical coherence tomography; landmark registration; coronary artery stenting

Domaines: Traitement des images [eess.IV]; Imagerie médicale

Laurent Sarry, Romain Dumas, Florian Dubuisson, Emilie Pery, Nicolas Combaret, et al.. Landmark-based registration of OCT image sequences for the follow-up of stent coverage and apposition. First International MICCAI-Workshop on Computer Assisted Stenting, Oct 2012, Nice, France. ⟨hal-02307931⟩