Low rank based FAZ segmentation in OCTA images.

Optical Coherence Tomography Angiography (OCTA) is a non-invasive method for vascular imaging of different tissues. Several diseases can disturb the blood supplying of retinal tissues which leads to loss of blood vessels and a reduction in the vascular density of retinal tissue. This also changes the size and shape of the FAZ (Fovea Avascular Zone) in the retinal OCTA images. Considering this fact, FAZ segmentation and analyzing its shape, can be used for diagnosing of several eye-related diseases. This paper is devoted to propose new methods for accurate segmentation of FAZ from OCTA images. The proposed methods have three steps: first, FAZ localization, for which, two methods have been proposed. The first one used Low-rank Tensor Ring (TR) decomposition of OCTA images and the other one exploited morphological operators. In the second step, the de-noised OCTA image is used for FAZ segmentation exploiting the information of FAZ localization, and in the third step, the resulting segmented FAZ is refined to derive a clean area. To the best of our knowledge, this is the first time of using low-rank TR estimation for FAZ segmentation. Simulation results confirmed the performance of the proposed methods for an accurate FAZ segmentation.