Self Regulating Classification of Healthy and AMD Class Retinal OCT Images with Deep-Learning Algorithms

The Optical Coherence Tomography (OCT) images offer cross-sectional imaging of retina excluding the need of conventional method. It enables the diagnosis of various retinal diseases by non-invasive technique. Age-related Macular Degeneration (AMD) is among the major eye diseases that affect human vision. Early detection of AMD may help to prevent the vision loss. This work deals deep learning approach diagnosis of diseases based on classification of OCT images into healthy and AMD groups. The proposed model has three stages of operation: binary classification, feature extraction and image resizing. Various deep learning algorithms are tested for the selected images and EfficientNetV2S approach produces better response based on the parametric metrics with an accuracy of 0.95.