Section: New Results

Image-based modeling

Modeling vasculature for real time simulation

One of our objectives to benefit interventional neuroradiology is to offer a patient-based interactive simulator to the interventional radiologists. Our contributions address vasculature modeling from patient data, namely 3D rotational angiography (3DRA) volumes. During Ahmed YUREIDINI's PhD thesis (2010-2014), a new model was developed consisting of a tree of local implicit blobby models.

We've been collaborating with SHACRA Inria project-team (Lille-Nord Europe) and the Department of Interventional Neuroradiology from Nancy University Hospital, in the context of the SOFA-InterMedS Inria Large-Scale Initiative. Ahmed YUREIDINI defended his PhD thesis in May this year with highest honors [9] . In particular, a detailed study was made to compare our tree of local implicits with triangular meshes in a view to model synthetic shapes as well as vasculatures from patient data. Increased performances with regard to processing speed, numerical stability and realism of the behavior were demonstrated.

Tools reconstruction for interventional neuro-radiology

Minimally invasive techniques impact surgery in such ways that, in particular, an imaging modality is required to maintain a visual feedback. Live X-ray imaging, called fluoroscopy, is used in interventional neuroradiology. Such images are very noisy, and cannot show but the vasculature and no other brain tissue. In particular, since at most only two projective fluoroscopic views are available, containing absolutely no depth hint, the 3D shape of the micro-tool (guidwire, micro-catheter or micro-coil) can be very difficult, if not impossible to infer, which may have an impact on the clinical outcome of the procedure.

In collaboration with GE Healthcare, we aim at devising ways to reconstruct the micro-tools in 3D from fluoroscopy images. Charlotte Delmas has been working as a PhD Cifre student on this subject since April 2013. A solution in a two-view reconstruction context was proposed this year based on the extraction of the guide-wire as a skeleton in the images. The large stereo basis (views are almost orthogonal) and the segmentation errors (such as both missing parts and spurious segments in the skeleton) make the reconstruction especially difficult. The skeletons are subdivided in simple curves that are matched to build all corresponding potential 3D curves. These curves are nodes in a graph whose edge weights express a connection cost that takes into account both distance and orientation at the curves extremities. The solution 3D curve is provided by following the path of minimal cost in the graph. This algorithm demonstrated very good reconstruction results on synthetic and phantom data. A paper on this subject has been accepted for publication at SPIE Medical Imaging 2015.

Patient-specific heart valve modeling

Many pathologies damage heart valve anatomy producing undesired backflow, or regurgitation, decreasing cardiac efficiency and potentially leading to heart failure if left untreated. Such cases could be treated by surgical repair for the valve. However it is technically difficult and outcomes are highly dependent upon the experience of the surgeon: he must essentially predict the displacement and deformation of complex valve leaflets and supporting structures. One way to facilitate the repair is to simulate the mechanical behavior of the pathological valve with patient-specific data. This is the objective of Pierre-Frédéric Villard’s one-year CNRS delegation in the Harvard Bio-robotics Laboratory (HBL). During the initial three first months of the sabbatical leave, various tasks have been performed: i) Study of the physiology of pathological valve behavior with medical experts. Following anatomical book reading and medical expert interviews the anatomy and the physiology are now understood. ii) Evaluation of HBL material for 4D ultrasound segmentation. HBL has previously developed a method to extract mitral valve geometry from a home-made high temporal resolution 3D ultrasound and iii) Automatic segmentation of a Mitral Valve microCT to feed a biomechanical model. A method to semi-automatically segment the leaflet-chordae set has been developed.