Automated Diagnosis Model for Glaucoma Detection: A Deep Learning Feature Fusion and LS-SVM based Approach

Glaucoma is an ocular disorder that affects millions of people and one of the second leading diseases in worldwide. Early detection of glaucoma is vital for avoiding irreversible vision loss problem. Our proposed model, based on deep learning-based ensemble pre-trained models like, Inception V3, GoogLeNet and VGG16 are compared with LS-SVM classifier to achieved better classification result (glaucoma or healthy) using G1020 dataset. Due to limited medical images in our dataset, we have applied data augmentation method is used to enhance the training size of the images in our preprocessing part. In our experimental result, we have observed that deep CNN-based ensemble pre-trained models with modified LS-SVM (least squares support vector machine) performed better classification result 93.90% using G1020 dataset as compared to other traditional models.