Automated Detection of Glaucoma in Fundus Images using Variational Mode Decomposition Textural Features

摘要

A variational mode decomposition (VMD) & local binary patterns (LBP) based features extraction from digital fundus images is proposed for glaucoma detection. The band-limited intrinsic mode images (BLIM:s) obtained by VMD, encompasses the varying spectral content embodying the non-linear and spatial non-stationary textural modulations in the fundus images. LBP feature descriptors apprehend the topographic tortuousness of the optical tissue fluids and substantiate the perturbations in intraocular fluid pressure (IOP) within the human eye which is caused due to glitches in the optical drainage system. Using artificial neural network, a classification accuracy of 95.2% is obtained on publicly available Medical Image Analysis Group (MIAG) dataset, which validates the suitability of the proposed framework in glaucoma identification.