Analysis of a Novel Segmentation Algorithm for Optical Coherence Tomography Images Based on Pixels Intensity Correlations

Abstract

One of the newest important medical imaging modalities is optical coherence tomography (OCT) providing the possibility of taking pictures from optical scattering tissues such as retina. With the help of accurate verification and analysis of OCT images, it is possible to identify and treat irreversible retinal diseases like glaucoma. Therefore, to suggest novel high-performance segmentation methods is of significant importance. In this article, a novel algorithm is proposed for the segmentation of OCT B-scans. The proposed method uses the intensity information of pixels to find a distinguishing feature for boundary pixels which are located on the retinal layer boundaries. Also, a nonparametric mathematical model is provided and analyzed for the segmentation algorithm. Such a novel model is also capable of determining boundary pixels. The performance of the proposed algorithm is evaluated for both normal and age-related macular degeneration (AMD) cases. The comparisons are performed from complexity and accuracy points of view. We show that the proposed algorithm has an outstanding performance in terms of values of mean signed and unsigned errors for both cases. Also, the results of mathematical analysis and execution of the algorithm match with good approximation.