Automated classification of retinal optical coherence tomography volumes based on RepVGG: Experimental validation for detection of dry age-related macular degeneration and diabetic macular edema

We introduce an algorithm for the automated classification of optical coherence tomography volumes based on the RepVGG artificial neural network. The algorithm is intended for the detection of dry age-related macular degeneration and diabetic macular edema from optical coherence tomography volumes. RepVGG is a simple yet powerful convolutional neural network that we use to classify optical coherence tomography volumes. Optical coherence tomography volumes from dry age-related macular degeneration, diabetic macular edema, and normal samples can be classified. We experimentally evaluated the effectiveness of the proposed algorithm on an optical coherence tomography dataset collected from 45 subjects (15 healthy subjects, 15 patients with dry age-related macular degeneration, and 15 patients with diabetic macular edema). The proposed algorithm correctly classified 73.3% of patients with dry age-related macular degeneration, 93.3% of patients with diabetic macular edema, and 86.7% of normal subjects. This algorithm may contribute to standardizing remote ophthalmic diagnosis of retinal diseases.