Hervé Delingette

Volumetric medical images segmentation using shape constrained deformable models

In this paper we address the problem of extracting geometric models from low contrast volumetric images, given a template or reference shape of that model. We proceed by deforming a reference model in a volumetric image. This reference deformable model is represented as a simplex mesh submitted to regularizing shape constraint. Furthermore, we introduce an original approach that combines the deformable model framework with the elastic registration (based on iterative closest point algorithm) method. This new method increases the robustness of segmentation while allowing very complex deformation of the original template. Examples of segmentation of the liver and brain ventricles are provided.

Virtual Reality, Augmented Reality and Robotics in surgical procedures of the liver

Medical image processing led to a major improvement of patient care. The 3D modeling of patients from their CT-scan or MRI thus allows a better surgical planning. Simulation offers the opportunity to train the surgical gesture before carrying it out. And finally, augmented reality provides surgeons with a view in transparency of their patient which, in the future, will allow to obtain automation of the most complex gestures. We present our latest results in these different fields of research applied to surgical procedures of the liver.

A New Kinetic Modeling Scheme for the Human Left Ventricle Wall Motion with MR-Tagging Imaging

Magnetic resonance tagging has proved to be an efficient non-invasive imaging technique for the study of the heart motion, producing a sharp and dense pattern on the tissues. Up to now, the quality of information derived from it has been unequalled by other modalities. It seems to be the perfect modality for the building of heart motion models. However, its main drawback was the tediousness of the image processing tasks it requires to extract the information. Recent achievements [1,9] have overcome that. In previous work [2], we have presented a novel parametric class of deformation for the 2D heart motion in the short axis planes. From it, we can compute a very accurate and compact analytical expression of the walls motion from the grid information. We present here this kinetic model, and its applications to heart modeling. In particular, we build a decomposition basis of the 2D motion with only three orthogonal modes.

From Semantics to Computer Science: Asclepios: a research project team at INRIA for the analysis and simulation of biomedical images

Asclepios1 is the name of a research project-team o cially launched on November 1st, 2005 at INRIA Sophia-Antipolis, to study the Analysis and Simulation of Biological and Medical Images. This research project-team follows a previous one, called Epidaure, initially dedicated to Medical Imaging and Robotics research. These two project teams were strongly supported by Gilles Kahn, who used to have regular scienti c in- teractions with their members. More generally, Gilles Kahn had a unique vision of the growing importance of the interaction of the Information Technologies and Sciences with the Biological and Medical world. He was one of the originators of the creation of a speci c BIO theme among the main INRIA research directions, which now regroups 16 di fferent research teams including Asclepios, whose research objectives are described and illustrated in this article.

Teaching tool for advanced visualization of temporal bone structures by fusion of μCT and CT scan images

Main goal nImprove the understanding of human temporal bone computed tomography (CT) scans based on semi-automatically segmented microcomputed tomography (µCT). n nIntroduction nThe three-dimensional ear anatomy is complex and challenging to interpret in CT scans because small structures are partially visible. Histological slices provide complementary high-resolution information, but may lead to geometrical distortions of the anatomy during preparation. Conversely, µCT preserves the shape. n n3D images acquisition and segmentation no Five freshly cadaveric pairs of temporal bones no Acquisition of CT (General Electric; Light Speed VTC 64) ) and a µCT (General Electric; eXplore speCZT) images no Seed-based segmentation of every reliable anatomical structures on CT or µCT n nRigid registration of µCT and CT no First, rough point-based registration using anatomical landmarks no Second, automatic rigid registration using a block matching framework n nResults nHigh resolution structures are fused and visualized on corresponding CT images. This experience significantly improves the visual recognition and spatial understanding of partially visible structures (e.g. tympanic scala, facial nerve and chorda tympani) in CT images. n nConclusion nGeometrically accurate temporal bone reconstructions provide an advanced teaching tool for medical students and cochlear implant surgeons. The understanding of spatial relationship between anatomical structures as well as the virtual exploration of surgical approaches is greatly facilitated.