Intraretinal fluid map generation in optical coherence tomography images

Abstract

[Abstract]: The retina represents one of the most studied parts of the human body, thanks to its easy access for any ophthalmological study. It represents the main neurosensory part of the eye and, by its study, pathologies not only from the visual system but also from different body systems (like the neural system or vascular system) can be detected. Among the principal medical imaging modalities that are used to study the eye fundus and the retinal histological structures, the optical coherence tomography (OCT) has proven to be one of the most reliable techniques. It is widely used for the detection of diseases like the diabetic macular edema and the age-related macular degeneration, which cause fluid leakages between the retinal layers. Both these diseases are among the main causes of blindness in developed countries, as these leakages tear the retinal structure and provoke the formation of other pathological bodies. As the symptoms start to be noticeable only when they cause a severe amount of damage to the retinal structures that may only be solved with surgical procedures, an early detection of these cystoid bodies became an impending matter. If these leakages are found in time, a simple and noninvasive treatment is sufficient to prevent further damage and reduce the damage already suffered.

As both pathologies are among the main causes of blindness, several studies have recently proposed methodologies to aid in their early detection and diagnosis. These proposals all circle around the same idea: generating a precise segmentation of these cystoid fluid leakages in the OCT images. These proposals have achieved an acceptable rate of success, but they limit their scope to perfectly define cystoid bodies. In reality, these cystoid bodies may appear mixed with normal retinal tissues and other pathological structures, not having a clear defined boundary to segment. Thus, an alternative paradigm for these fluid leakage detections was proposed, based on a regional analysis and an intuitive robust representation. In this chapter, we explain and provide an overview of this paradigm, and possibilities for its implementation in computer-aided diagnostic systems.