Current research on glaucoma detection using compact variational mode decomposition from fundus images

Abstract

Glaucoma is one of the foremost causes of blindness over the world. It develops slowly and damages optic nerve head. Existing methods of glaucoma detection are expensive and sluggish. Hence quick and low-priced methods are required. In this paper, a novel fully variational and adaptive computer based glaucoma detection using compact variational mode decomposition (CVMD) from fundus images is proposed. Efficient sub band images having narrow fourier bandwidth, clear and sharp boundaries are obtained using CVMD is the key idea of the proposed method. This gives robust estimate of features. Texture feature capture subtle variation and give more information therefore these features are extracted from efficient sub band images and helps to increase the glaucoma detection accuracy. Extracted features are normalized and classified using support vector machine classifier. The obtained classification accuracies are 86.67 %, 86.67 %, 85.42% and 89.18 % for three, five, eight and tenfold cross validation respectively with kernel parameter 2. The obtained accuracy, sensitivity, specificity, precision and Fmeasure are 89.18 %, 90 %, 85 %, 93.34 % and 89.34 % respectively for tenfold cross validation. Proposed method is found to be better compared to the existing.