| Symptomatic exudate-associated derangements (SEAD) associated with abnormalities in the retina, such as age-related macular degeneration (AMD), this paper proposed automatic segmentation algorithms of SEAD based on some medical image segmentation algorithms for spectral domain optical coherence tomography (SDOCT) retinal images. In this paper, the main work and research includes the following several aspects.(1) Image preprocessing. Due to the influence of speckle noise, blood vessels, movement of sensors when imaging and the light conditions, there are a lot of speckle noise in SDOCT retinal images, so the image quality is very poor. This paper made image preprocessing with bilateral filter to improve the image quality.(2) Segmentation of internal limiting membrane (ILM) and retinal pigment epithelium (RPE). This paper analyzed the ILM features of different images, and got the better ILM boundary with a threshold method. At the same time, this article automatically segmented the RPE boundary with graph theory and dynamic programming. This algorithm got the better RPE boundary using the high reflection feature of RPE layer and the prior information of retina.(3) Automated segmentation of SEAD based on clustering. Firstly, the algorithm utilizes K-means clustering method to get the rough SEAD position, and then uses the detected blood vessel position information, the spatial information and gray information of SDOCT retinal images to remove the incorrect segmentation, and finally gets more accurate SEAD segmentation result.(4) Automated segmentation of SEAD based on level set. The algorithm utilizes threshold segmentation method to get the initial contour curve of level set, and then adopts the level set method to get the rough SEAD position. Following these, we use the detected blood vessel position information, the spatial information and gray information of SDOCT retinal images to remove the incorrect segmentation. Finally more accurate SEAD segmentation results can be obtained. |
PDF Full Text Request