Discovery of knowledge patterns in retinal images for identification of pathologies through image processing and data mining techniques

Abstract

Retinal image analysis has evolved rapidly to comprehend the subtle changes of the retina suitable for retinal disease diagnosis. There are different imaging modalities, namely Optical Coherence Tomography, Hyper Spectral Imaging, Fundus photography, etc., to get the fine details of the retina. Among them, Fundus photography is the most commonly used retinal image screening technique for the identification of pathology. It can be broadly classified into retinal anatomical structural patterns that capture the anatomical details of the retina and retinal lesion patterns, which exposes bright and dark pathological distractions caused due to retinal abnormalities. Recently, computer-aided retinal image analysis has been developed by researchers that can retrieve delicate information from the fundus images used to identify retinal diseases that might not have been seen from the manual investigation. For the analysis of retinal anatomical structural patterns, retinal lesion pattern, disease classification, as well as image quality categorization, computational methods are employed. newlineThis research has focused on retinal image analysis in three ways: extraction of retinal anatomical structural and lesion pattern, retinal disease classification and fundus image quality categorization from retinal fundus images. Fourteen publicly available retinal fundus image benchmark datasets and an image feature dataset were obtained for this research work. They are newline newline