Accurate Segmentation of Arterial Walls in Adaptive Optics Images

In this paper, a new approach is presented for accurately delineating the artery walls in adaptive optics images. To the best of our knowledge, this is the first paper addressing this problem in this context. To achieve this goal, we propose an automatic two-steps process. The position of inner walls of each vessel is first roughly estimated. Then, a deformable model embedding a parallism constraint is used to refine their positioning and localize the outer walls. Such an approach enables us to model the walls as approximate parallel contours and to control their distance with respect to their common reference line. This model leads to four independent energy functionals, minimized by gradient descent. Noticeably, the evaluation of this new model on healthy subjects shows an overall error smaller than the inter-physicians variability.