Automatic detection of blood vessels in retinal OCT images

Abstract

The eye is a non-invasive window where clinicians can observe and study in vivo the retinal vasculature, allowing the early detection of different relevant pathologies. In this paper, we present a complete methodology for the automatic vascular detection in retinal OCT images. To achieve this, we analyse the intensity profiles between representative layers of the retina, layers that are previously segmented. Then, we propose the use of two threshold-based strategies for vessel detection, a fixed and an adaptive approach. Both methods have been tested and validated with 128 OCT images, that include 560 vessels that were labelled by an ophthalmologist. The approaches provided satisfactory results, facilitating the doctors’ work and allowing better analysis and treatment of vascular diseases.