An Automatic Fluid Segmentation For Retinal SD-OCT Images With Active Contour Model

Optical coherence tomography(OCT)is becoming an increasingly important modality for the diagnosis and management of a variety of eye diseases,such as central serous chorioretinopathy(CSC).Spectral domain OCT(SD-OCT),an advanced type of OCT,produces three dimensional high-resolution cross-sectional images and demonstrates delicate structure of the functional portion of posterior eye,including retina.As the clinical importance of SD-OCT for retinal disease management and the need for quantitative and objective disease analysis grows,fully automated subretinal fluid segmentation algorithm is required.In this thesis,unsupervised and supervised fluid segmentation algorithms with two stages are proposed.In the first stage,the candidate fluid region is automatically estimated to obtain the initial curve of the fluid area for the level set method.In the second stage,the local Gaussian pre-fitting energy model and the linear discriminant analysis based fitting energy are proposed to segment subretinal fluid.The testing data set with 23 longitudinal SD-OCT cube scans from 12 eyes of 12 patients are used to evaluate the proposed algorithms.Without retinal layer segmentation,the proposed algorithm can obtain high segmentation accuracy.Our model may provide reliable subretinal fluid segmentations for NRD from SD-OCT images and shows the potential to improve clinical therapy for CSC.