| Abstrakt: |
Worldwide, glaucoma is a type of eye disease, which causes loss of vision by damaging optic nerves within the eye. Available glaucoma detection techniques are less accurate. This paper proposed a computer-based glaucoma detection using second-stage quasi-bivariate variational mode decomposition (SS-QB-VMD)-based fine sub-band images (SBIs) from fundus images. The preprocessed images are decomposed into five SBIs at the first stage using QB-VMD. The high-frequency fifth SBI obtained using the first stage is further decomposed into five SBIs using QB-VMD at the second stage. At the second stage, the decomposed SBIs are more discriminating than obtained at the first stage. These decomposed SBIs at second stages are fine with no mode mixing problems. These are finer and help to mine more useful detailed information to detect glaucoma more accurately. Texture features are extracted from decomposed SBIs. Extracted texture features are normalized using the z-score method and classified using the support vector machine classifier. The obtained accuracy, sensitivity, specificity, precision, dice value, and Jaccard index value are 92.67%, 91.43%, 93.75%, 94.03%, 91.98%, and 86.1%, for tenfold cross-validation, respectively. The results demonstrated that the proposed method is better compared to the existing methods for same images. It can be used to detect glaucoma more accurately. [ABSTRACT FROM AUTHOR] |
