Ophthalmology Disease Relationships Based on Deep Metric Learning.

This study addresses the development of a deep metric learning model to detect early retinal abnormalities those are generally considered to precede ocular diseases or systemic diseases. By mapping retinal images into a two-dimensional feature space reflecting disease similarity, our model identifies subtle deviations not easily detectable through conventional methods, using the RFMiD dataset of approximately 3,200 images across 45 disease and normal states.