Real time data fusion for bone tracking in CAOS
In the context of Computer-Assisted Orthopaedic Surgery, bones are generally tracked by means of an implanted rigid body localized for instance by an optical sensor. We have proposed to replace this invasive method by a 3D ultrasound-based tracking of the bone. One difficulty of the approach lies in the specificity of ultrasound images of the bone (cf. figure). After a first approach developed during the PhD of J.Schers, we started a new project. In a first time, a robust registration method to map CT planning to the US intra-operative conditions has been developed in the context of hip surgery. This method is a variant of ICP which proposes, among other features, the automatic and robust initialization of the searched transform. It has been tested for US/CT and US/US registration on sawbones and accuracy and robustness are very encouraging. We also proposed specific clinically relevant metrics for accuracy evaluation
References: Haddad, O., Leboucher, J., Troccaz, J., & Stindel, E. (2016, December). Initialized Iterative Closest Point for bone recognition in ultrasound volumes. In Pattern Recognition (ICPR), 2016 23rd International Conference on (pp. 2801-2806). (PDF) O.Haddad, J.Leboucher, J.Troccaz, E.Stindel. Robust rigid registration for non invasive Computer Assisted Orthopedic Surgery (CAOS). Preliminary results. Proceedings of IEEE ISBI’2015, New-York, April 2015 (PDF)
Collaborations: LATIM laboratory in Brest since 2013.
Status: ACTIVE Two PhD thesis defended. The first one (J.Schers, 2009) in collaboration with Praxim is associated to a patent. After a standby phase, this project started again in the fall 2013 with a new thesis (prepared by Oussama HADDAD) launched in collaboration with Pr Eric Stindel (LATIM, orthopedic surgeon, Brest).
Contact: Jocelyne.troccaz@univ-grenoble-alpes.fr or Eric.Stindel@univ-brest.fr