Automatic System for the Identification and Visualization of the Retinal Vessel Tree Using OCT Imaging ^†

Abstract

This paper proposes a system for the three-dimensional identification and visualization of the retinal vasculature using Optical Coherence Tomography (OCT) scans. This fully automatic tool provides useful biomarkers to the medical specialists that facilitate the prevention, diagnosis and treatment of various retinal and systemic pathologies.

1. Introduction

The eye fundus is the only part of the human body where the blood vessels can be directly visualized non-invasively and safely in real time. An accurate analysis of the morphology of the retinal microvasculature allows the early diagnosis of different systemic diseases that can affect multiple parts of the body, such as hypertension, diabetes or arteriosclerosis, all of which are leading global public health concerns.

Nowadays, Optical Coherence Tomography (OCT) technique is increasing its use in clinical studies, health service research and daily clinical practice. This biomedical imaging system is capable of providing high-resolution cross-sectional scans of the internal microstructure of different retinal tissues in real time. These OCT scans allow the medical specialists to analyze and identify anatomical and physiological changes in the complex microvascular network of the retina.

In this work, we propose the development of a novel and complete methodology for the automatic identification and three-dimensional visualization of the retinal vasculature using OCT scans. This fully automatic system provides an intuitive visualization tool that facilitates a more complete and reliable analysis of the complex vascular structure of the retina.

2. Methodology

The proposed methodology is divided into four main steps: a first step, where we extract the two-dimensional vessel tree in the NIR retinography image; second, we estimate the vessel calibers; third, we obtain the corresponding depth of the vessel profiles in the OCT sections; and a final fourth step, where we make a complete three-dimensional representation of the retinal vasculature. This system allows the manipulation of the retinal vessel tree by means of graphical transformations including translation, scaling and rotation and their arbitrary combinations [1].

3. Results and Conclusions

The proposed method reached satisfactory results, providing a consistent and coherent three-dimensional retinal vessel tree reconstruction and visualization that can be posteriorly used in different medical analysis and diagnostic processes of various retinal and systemic pathologies. Figure 1 shows a representative example of the three-dimensional reconstruction and visualization where the caliber scale bar is also presented in the screen to facilitate the analysis of the medical specialists.