Interretinal Symmetry in Color Fundus Photographs

retina, fundus, symmetry, biometrics

Symmetry can be defined as uniformity, equivalence or exact similarity of two parts divided along an axis. While our left and right eyes clearly have a high degree of external bilateral symmetry, it is less obvious to what degree they have internal bilateral symmetry. In this paper, we try to find approximate-bilateral symmetry in retina, one of the internal parts of our eye, which plays a vital role in our vision and also can be used as a powerful biometric. Contrary to previous works, we study interretinal symmetry from a biometric perspective. In other words, we study whether the left and right retinal symmetry is strong enough to reliably tell whether a pair of the left and right retinas belongs to a single person. For this, we focus on overall symmetry of the retinas rather than specific attributes such as length, area, thickness, or the number of blood vessels. We evaluate and analyse the performance of both human and neural network based bilateral retina verification on fundus photographs. By experimenting on a publicly available data set, we confirm interretinal symmetry.