The Research On OCT Retinal Layer Segmentation Based On Groupwise Curve Alignment

Retinal layer's thickness measurements have been shown to correlate well with the severity of different ocular diseases;hence,they provide useful diagnostic information concerning diseases.Retinal layer segmentation from optical coherence tomography(OCT)is of fundamental importance for measuring retinal layer thicknesses.Manual segmentation of retinal layers from OCT remains dominant in ophthalmological clinical practice but has serious drawbacks: it is time consuming,labor intensive and results in inter/intra-rater variations.Computer aided segmentation has attracted intensive research attention because it holds the potential not only to provide repeatable,quantitative and objective results but also to reduce the time and effort required to delineate the retinal layers.However,most of the existing computer based retinal layer segmentation techniques focus on segmenting specific layers by exploring their unique characteristics;thus,they can fail to segment a retinal layer that is totally different.In this paper,we propose a generative retinal layer segmentation method based on group-wise curve alignment that combines the capabilities of segmenting different retinal layers into a unified framework.This method is unique for both its accuracy and its ability to segment any retinal layer without any special modifications.Our method is potentially useful in a large variety of practical applications involving retinal layer segmentation from OCT.The main contributions of this paper are as follows.(1)We present an early work on group-wise curve alignment and propose a generative retinal layer segmentation model based on group-wise curve alignment.The shapes of gray-value curves of the same OCT retinal image are similar because of the same sequence of tissue structure from top to bottom corresponding to each column.This work provides a possibility of OCT retinal layers segmentation on retinal images.(2)We present an early work on semi-automated segmentation of retinal layers from OCT that can delineate the boundaries of any retinal layer after a single boundary point is manually specified.Our framework of multilayer segmentation also can segment any stratified structure that caused by disease without any special technical re-design or experimental retraining.(3)We propose an alternative strategy for retinal layer segmentation from OCT that involves a group-wise curve alignment.This strategy avoids the redundancies of the pair-wise curve alignment and provides an opportunity to correct the errors in pairwise maps using a constrained combinatorial optimization process.(4)We experimentally validate that the proposed method outperforms a representative state-ofthe-art technique by using images of both two-dimensional and three-dimensional.