Compact Scattering Features for Glaucoma Detection

Abstract In the field of medical image diagnostics, automatic glaucoma classification is a daunting task that requires carefully crafted features (Machine Learning) or tuning hyperparameters over huge search space (Deep Learning). this paper, we proposed wavelet scattering based approach extracts from fundus images at different sub-stages, reduces feature using principal component analysis (PCA) and represents classifying in terms projection loss along subspace for each class. This simple yet elegant yields fast-paced results with great interpretability, deep learning-based approaches lack, excels most performance metrics when compared to other machine learning methods.