Geometric deformable model driven by CoCRFs: application to optical coherence tomography.

We present a geometric deformable model driven by dynamically updated probability fields. The shape is defined with the signed distance function, and the internal (smoothness) energy consists of a C1 continuity constraint, a shape prior, and a term that forces the zero-level of the shape distance function towards a connected form. The image probability fields are estimated by our collaborative Conditional Random Field (CoCRF), which is updated during the evolution in an active learning manner: it infers class posteriors in pixels or regions with feature ambiguities by assessing the joint appearance of neighboring sites and using the classification confidence. We apply our method to Optical Coherence Tomography fundus images for the segmentation of geographic atrophies in dry age-related macular degeneration of the human eye.