Computer Vision News - February 2022

46 Congrats, Doctor! as a key modality for this investigation. OCT-A is a fast and dyeless technique that allows the visualisation of the microvasculature in vivo and to the deepest level of detail. Nonetheless, the potential of this technology in clinical diagnostics is still under investigation. In this research, we develop a fully automated framework for the analysis of OCT-A scans, we propose novel retinal measurements to characterise the microvasculature in the eye, and we show the application of our framework to a wide spectrum of conditions with the aim of demonstrating the full potential of this technology. OCT-A computational framework The first step of the OCT-A computational framework is imageprocessing. SinceOCT-A is a fairly recent technology, standard blood vessel segmentation has not yet been established. Hence, we first explored the best image processing pipeline by creating an original OCT-A dataset with manually segmented scans (publicly available) by investigating a large set of blood vessel segmentation methodologies. Handcrafted filters and neural network architectures were surveyed and evaluated using standard performance metrics and newly customised OCT-A vascular measures. Results showed the superiority of deep learning approaches over handcrafted filters and the susceptibility of clinical measurements toeachof the segmentation approaches , suggesting the need to pay special attention to the image processing step when performing meta-analyses. ThesecondstepoftheOCT-Acomputational framework is conducting network analysis for the discovery of retinal biomarkers in the segmented OCT-A images. Common Ylenia Giarratano has recently completed her PhD in Precision Medicine at the University of Edinburgh. Her research interests focus on retinal imaging, computational modeling, and machine learning with the aim of revealing retinal biomarkers of microvascular diseases. She is currently pursuing her work at the University of Edinburgh as a research fellow, investigating retinal microvascular predictors of compromised brain haemodynamics in small vessel disease. Congrats, Doctor Ylenia! Studies over recent years have reported that changes in the retinal vasculature due to disease may happen years before symptoms manifest. There has thereby been an increased interest in using retinal imaging for the discovery of biomarkers of both ocular and systemic disease. Among retinal imaging modalities, optical coherence tomography angiography (OCT-A), a recent technology, has emerged