Computer Vision News - February 2022

47 Ylenia Giarratano studies with a retinal vascular footprint: diabetic retinopathy (DR), chronic kidney disease (CKD), and living kidney donation. In the DR study, we were able to reproduce previously reported results on changes at the microvascular level and achieve good performances in classifying patients with DR. Whereas changes in the eye of patients with CKD were more subtle and challenging to detect. Finally, despite living kidney donors being considered near- healthy patients, they remain at higher risk of developing cardiovascular disease and CKD. We investigated, for the first time, the possibility of OCT-A retinal imaging as a tool to improve the targeting of patients at risk. Our results suggested that OCT-A microvascular phenotypes may provide further insights into the long-term risk assessment of kidney donation. retinal characteristics such as vessel density, vessel radius, and vessel tortuosity were computed based on current clinical knowledge. In addition, we developed novel microvascular metrics based on geometrical, topological, and functional properties of the vascular network to cover the full spectrum of possible retinal measurements, enabling the hypothesis- free discovery of new clinically relevant biomarkers of diseases (see Figure with OCT-A computational framework). Applications The proposed candidate retinal biomarkers computed by our OCT-A framework can be used to explore associations in ocular and systemic dysfunctions and to build machine learning classifiers for the prediction of patient status. We have demonstrated the application of the OCT-A computational framework in three case

RkJQdWJsaXNoZXIy NTc3NzU=