Region-segmentation strategy for Bruch’s membrane opening detection in spectral domain optical coherence tomography images

• DOI: 10.1364/boe.10.000526
• Publication Date: 2019-01-14T22:28:42Z
• Authors: Zailiang Chen; Peng Peng; Hailan Shen; Hao Wei; Pingbo Ouyang; Xuanchu Duan
• Keywords: Bruch's membrane

Bruch's membrane opening (BMO) is an important biomarker in the progression of glaucoma. minimum rim width (BMO-MRW), cup-to-disc ratio spectral domain optical coherence tomography (SD-OCT) and lamina cribrosa depth based on BMO are measurable parameters for glaucoma diagnosis. The accuracy measuring these significantly affected by detection. In this paper, we propose a method automatically detecting SD-OCT volumes accurately to reduce impact border tissue vessel shadows. includes three stages: coarse detection stage composed retinal pigment epithelium layer segmentation, optic disc multi-modal registration; fixed U-net which transformed into region segmentation problem area bias component proposed loss function; post-processing consistency results remove outliers. Experimental show that outperforms previous methods achieves mean error 42.38 μm.