Automated segmentation of fluid regions in optical coherence tomography B-scan images of age-related macular degeneration

Abstract Age-related macular degeneration (AMD) is a common eye disease that causes progressive vision loss in people older than 50 years. Fluid regions in retina are the most characteristic of AMD. Accurately segmenting fluid regions is crucial for the early diagnosis of AMD, and assessment of treatment efficacy. In this paper, we propose an automatic deep learning method constructed by integrating Squeeze-and-Excitation blocks with U-Net named SEUNet to segment fluid regions and classify OCT B-scan images to AMD or normal image. The proposed method comprises three stages: (1) preprocessing stage that includes image noise removal, locating the image on the area of interest, and image color-reversing; (2) fluid region segmentation stage which is based on U-Net and constructed by integrating Squeeze-and-Excitation block to segment fluid region; and (3) image classification stage that classifies image to AMD or normal image. Experimental results show that the proposed method have an average IOU coefficient of 0.9035, an average Dice coefficient of 0.9421, an average precision of 0.9446, and an average recall of 0.9464. Therefore, the proposed method can effectively segment fluid regions in OCT B-scan images.