Section: New Results

Graph-based registration and segmentation

Paticipants: Enzo Ferrante, Vivien Fecamp, Aimilia Gastounioti, Bharat Singh, Stavros Alchatzidis, Nikos Paragios

Deformable image registration plays a fundamental role in many clinical applications. We investigated the use of graphical models in the context of a particular type of image registration problem, known as slice-to-volume registration. We introduced a scalable, modular and flexible formulation that can accommodate low-rank  [5] and high order  [16] terms, that simultaneously selects the plane and estimates the in-plane deformation through a single shot optimization approach. We applied our models on simulated and real-data in the context of ultrasound and magnetic resonance registration, demonstrating the potential of our methods.

We also developped a novel methodology for graph-based motion-driven segmentation  [24] and applied it for carotid plaque segmentation in ultrasound images. We identified the plaque region by exploiting kinematic dependencies between the atherosclerotic and the normal arterial wall. The methodology exploits group-wise image registration towards recovering the deformation field, on which information theory criteria are used to determine dominant motion classes and a map reflecting kinematic dependencies, which is then segmented using Markov random fields.

Moreover, in order to address the problem of general purpose multi-modal deformable registration/fusion we developped a novel and robust method using a metric defined in an appropriate sub-space which is adaptive to the image-content/image-modality  [18] . We adopted a graph-based formulation that assumes that intensities of corresponding pixels in the two image domains are related through an unknown piece-wise constant linear function. This relation is propagated to an appropriate sub-space (wavelets coefficients) where a criterion that couples the estimation of the deformation field with optimal transport function on the subspace and the smoothness of the deformation is considered.