Geographic Atrophy Segmentation for SD-OCT Images by MFO Algorithm and Affinity Diffusion

Age-related macular degeneration (AMD) is a common cause of vision loss among the elderly in developed countries. Geographic atrophy (GA) appears in advanced stages of non-exudative AMD. In this paper, we present a hybrid GA segmentation model for spectral-domain optical coherence tomography (SD-OCT) images. The method first segments the layered structure of the SD-OCT scan data and produces the projection images. Then we construct the histogram of the resulting image into a probability distribution function, and use this function to fit a Gaussian mixed model (GMM) by Moth-flame optimization (MFO) algorithm. To incorporate the globe spatial information to over come the impact of noise, a robust affinity diffusion method is proposed to construct the affinity map. Finally, bias field correction process is employed to remove the intensity inhomogeneity. Two data sets, respectively consisting on 55 SD-OCT scans from twelve eyes in eight patients with GA and 56 SD-OCT scans from 56 eyes in 56 patients with GA, are utilized to quantitatively evaluate the segmentation algorithm. Experimental results demonstrate that the proposed algorithm can achieve high segmentation accuracy.