Fovea-Attentive Residual Network for Extracting 3D Angiograms from Single Volumetric OCT Scan Peng Liu 1,2 , Jianlong Yang 1,* , Lixin Duan 2 , Yan Hu 1 , and Jiang Liu 1,3 1 Cixi Institute of Biomedical Engineering, Chinese Academy of Sciences, China * yangjianlong@nimte.ac.cn 2 Big Data Research Center at University of Electronic Science and Technology of China 3 Southern University of Science and Technology Abstract. Optical Coherence Tomography Angiography (OCTA) has drawn numerous attentions in ophthalmology. However, its data acqui- sition is time-consuming, because it is based on temporal-decorrelation principle thus requires multiple repeated volumetric OCT scans. In this paper, we propose a Fovea-Attentive Residual Network (FAR-Net). By combining an attention mechanism and residual learning, it is able to extract capillary-level angiograms directly from a single OCT scan. The segmentation results of the inner limiting membrane and outer plexiform layers and the central 1 ×1 mm 2 field of view of the fovea are employed in the fovea-attention mechanism. So the influences of large retinal vessels and choroidal vasculature on the extraction of capillaries can be mini- mized during the training of the FAR-Net. The residual learning is used to take advantages of the residual block’s shortcut mechanism, so as to avoid the gradient vanishing and accelerate the network convergence. The quantitative and qualitative results demonstrate that the FAR-Net has the capacity to better-visualizing capillaries around the foveal avascular zone than the existing work using a U-Net architecture. Keywords: Optical coherence tomography angiography · Capillary vi- sualization · Artificial intelligence. 1 Introduction Optical Coherence Tomography Angiography (OCTA) is an emerging functional imaging modality of OCT. In the last five years, it has drawn tremendous in- terests in clinical researches [1]. OCTA is capable of discriminating blood flow of retinal vessels with surrounding static tissue without the dye injection pro- cess, which is obligatory in traditional Fluorescein Angiography (FA). Besides, its capacity in resolving depth vascular information has shown superiority in the early detection and diagnosis of prevailing ocular diseases, such as diabetic retinopathy [2], age-related macular degeneration [3], and glaucoma [4]. However, existing OCTA techniques require much longer acquisition time than OCT, which brings difficulties to the eye fixation of the observed subject. arXiv:1906.07091v1 [physics.med-ph] 17 Jun 2019