Shape and texture based identification of glaucoma from retinal fundus images. (March 2022)

• Record Type: Journal Article
• Authors: Sonti, Kamesh; Dhuli, Dr. Ravindra
• Is Part Of: Biomedical signal processing and control. Volume 73(2022)
• Publication Date: 2022-03
• Subjects: Glaucoma -- Quasi bi-variate variational mode decomposition -- Pyramid histogram oriented gradient -- Invariant Haralick texture features -- Exponential polynomial support vector machines; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.103473
• Languages: English
• ISSNs: 1746-8094

Abstract

Graphical abstract: Highlights: Decompostion of images using QB-VMD technique. Extract shape and texture features using PHOG and Invariant Haralick Features. Classify the model with EP-SVM and Bagged ensemble approaches. Enhances the performance of the proposed model. Abstract: Among the eye disorders, glaucoma occurs due to the increase of intra-ocular pressure that causes irreversible damage of the optic nerve and leads to blindness. Therefore to avoid high cost machine usage, a novel method proposed for detecting glaucoma using shape and texture features. Initially, the retinal fundus images are decomposed using quasi bi-variate variational mode decomposition (QB-VMD) technique, the frequencies obtained from QB-VMD subjected to pyramid histogram oriented gradient (PHOG) and invariant Haralick texture features. The extracted combinational features are classified using combination of exponential polynomial support vector machines (EP-SVM) and bagged ensemble approach. The proposed method simulated on ACRIMA and Drishti-GS1 datasets using 10-fold cross validation and evaluated the performance metrics like accuracy, sensitivity, specificity and F-score. The simulation results and evaluation metrics show that the proposed approach achieves superior classification performance compared to other state-of-the-art approaches.