Fundus Image Quality Assessment Through Analysis of Illumination, Naturalness, and Structure Level

Abstract

AbstractOptical diseases such as diabetic retinopathy, cataract and glaucoma generally have affected a more quantity of the populace international. Image-based medical diagnosis relies upon adequate image quality and clarity. Mostly the fundus image acquisition are more vulnerable to distortions due to changes from a fixed place camera to portable fundus camera. The fundus images are the main tool for many retinal disease analyses. Repeated image acquisition is required for diagnosis because most of the images captured are of low quality. So, the automatic system is needed to evaluate the quality of the fundus image in terms of illumination level, naturalness level, and structure level. During the capturing time the Non-mydriatic image quality is more vulnerable to distortions. This kind of image quality distortions is called as generic quality distortions. The proposed work is about classification of the fundus image using the Adaptive Migration Biogeography based Optimization (AMBBO) Algorithm based Radial Basis Function Network (RBFN) where the Fundus Image Quality is assessed through the Analysis of Illumination, Naturalness, and Structure level (FIQAINS) model. Changes have been proposed to the original BBO algorithm since it does not have the intrinsic property of clustering which is necessary in fundus image classification. Hence the modified algorithm is used to classify the fundus image based on the generic illumination feature, the critical naturalness feature and the necessary structure feature. The experimental results that make use of MATLAB tools to accomplish the generic overall quality classification such as sensitivity, specificity, accuracy and AUC and different threshold values from DRIMDB Dataset of fundus images and furthermore discover accept or reject class based on the AMBBO Algorithm.KeywordsFundus imagesQuality distortionsIlluminationStructure levelNaturalnessAMBBO algorithm