Vessel Tree Extraction and Depth Estimation with OCT Images

Abstract

The identification of the retinal arterio-venular tree is a relevant issue for its analysis in a large variability of procedures. Classical methodologies employ 2D acquisition strategies that obtain a limited representation of the vascular structure. This paper proposes a new methodology for 2D vessel tree extraction and the corresponding depth estimation using Optical Coherence Tomography (OCT) images. This way, the proposal defines a more complete scenario for an adequate posterior vasculature analysis. The methodology employs different image analysis techniques to initially extract the 2D vessel tree. Then, the method maps these 2D positions in the corresponding histological sections of the OCT images and estimates the corresponding depths along all the vessel tree. To test and validate this proposal, this work employed 196 OCT histological images with the corresponding near infrared reflectance retinographies. The methodology provided promising results, indicating an acceptable accuracy in a complex domain as is the vessel tree identification. It provides a coherent 2D vessel tree extraction with the corresponding depth estimations that constitute a scenario with high potentially useful information for posterior medical analysis and diagnostic processes of many diseases as, for example, hypertension or diabetes.