2019/10/07 by Haris Cheong, Sripad Krishna Devalla, Cheong, Haris +21
Engineering · Medicine · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #Glaucoma and retinal disorders #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Optical Coherence Tomography Applications #Retinal Imaging and Analysis #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.1910.02844
openalex publication_date 2019/10/07 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Purpose: To remove retinal shadows from optical coherence tomography (OCT)\nimages of the optic nerve head(ONH).\n Methods:2328 OCT images acquired through the center of the ONH using a\nSpectralis OCT machine for both eyes of 13 subjects were used to train a\ngenerative adversarial network (GAN) using a custom loss function. Image\nquality was assessed qualitatively (for artifacts) and quantitatively using the\nintralayer contrast: a measure of shadow visibility ranging from 0\n(shadow-free) to 1 (strong shadow) and compared to compensated images. This was\ncomputed in the Retinal Nerve Fiber Layer (RNFL), the Inner Plexiform Layer\n(IPL), the Photoreceptor layer (PR) and the Retinal Pigment Epithelium (RPE)\nlayers.\n Results: Output images had improved intralayer contrast in all ONH tissue\nlayers. On average the intralayer contrast decreased by 33.7\±6.81%,\n28.8\±10.4%, 35.9\±13.0%, and43.0\±19.5%for the RNFL, IPL, PR, and RPE\nlayers respectively, indicating successful shadow removal across all depths.\nThis compared to 70.3\±22.7%, 33.9\±11.5%, 47.0\±11.2%,\n26.7\±19.0%for compensation. Output images were also free from artifacts\ncommonly observed with compensation.\n Conclusions: DeshadowGAN significantly corrected blood vessel shadows in OCT\nimages of the ONH. Our algorithm may be considered as a pre-processing step to\nimprove the performance of a wide range of algorithms including those currently\nbeing used for OCT image segmentation, denoising, and classification.\n Translational Relevance: DeshadowGAN could be integrated to existing OCT\ndevices to improve the diagnosis and prognosis of ocular pathologies.\n