Comprehensive retinal image analysis: image processing and feature extraction techniques oriented to the clinical task

Abstract

Medical digital imaging has become a key element of modern health care procedures. It provides a visual documentation, a permanent record for the patients, and most importantly the ability to extract information about many diseases. Ophthalmology is a field that is heavily dependent on the analysis of digital images because they can aid in establishing an early diagnosis even before the first symptoms appear. This dissertation contributes to the digital analysis of such images and the problems that arise along the imaging pipeline, a field that is commonly referred to as retinal image analysis. We have dealt with and proposed solutions to problems that arise in retinal image acquisition and longitudinal monitoring of retinal disease evolution. Specifically, non-uniform illumination, poor image quality, automated focusing, and multichannel analysis. However, there are many unavoidable situations in which images of poor quality, like blurred retinal images because of aberrations in the eye, are acquired. To address this problem we have proposed two approaches for blind deconvolution of blurred retinal images. In the first approach, we consider the blur to be space-invariant and later in the second approach we extend the work and propose a more general space-variant scheme. For the development of the algorithms we have built preprocessing solutions that have enabled the extraction of retinal features of medical relevancy, like the segmentation of the optic disc and the detection and visualization of longitudinal structural changes in the retina. Encouraging experimental results carried out on real retinal images coming from the clinical setting demonstrate the applicability of our proposed solutions.