Diabetic Macular Edema characterization by automatic analysis of Optical Coherence Tomography

Abstract

Diabetic Macular Edema (DME) is one of the most important complications of diabetes and a leading cause of preventable blindness in the developed countries. Among the di erent image modalities, Optical Coherence Tomography (OCT) is a non-invasive, cross-sectional and high-resolution imaging technique that is commonly used for the analysis and interpretation of many retinal structures and ocular disorders. In this way, the development of Computer-Aided Diagnosis (CAD) systems has become relevant over the recent years, facilitating and simplifying the work of the clinical specialists in many relevant diagnostic processes, replacing manual procedures that are tedious and highly time-consuming. This thesis proposes a complete methodology for the identi cation and characterization of DMEs using OCT images. To do so, the system combines and exploits di erent clinical knowledge with image processing and machine learning strategies. This automatic system is able to identify and characterize the main retinal structures and several pathological conditions that are associated with the DME disease, following the clinical classi cation of reference in the ophthalmological eld. Despite the complexity and heterogeneity of this relevant ocular pathology, the proposed system achieved satisfactory results, proving to be robust enough to be used in the daily clinical practice, helping the clinicians to produce a more accurate diagnosis and indicate adequate treatments