Proposal for Dry Eye Detection Caused by Contact Lenses Using a Smartphone with a Ring Light and Deep Learning Technology

Abstract

Using a smartphone, a ring light, and a grid-like cylinder, a method for detecting dry eye caused by contact lenses is proposed. When looking into a grid-like cylinder containing a ring light generated by a simple construction, a concentric grid appears to the eye, but dry eye during contact lens wear distorts this grid more significantly than the case for naked eye. Thus, multiple dry eye and nondry eye images were captured and used to train a neural network. When a new image was input to the trained neural network, the model gave a output indicating the likelihood of the subject having dry eyes. By analyzing the final convolution layers of the trained neural network, the dry eye characteristics were identified; this can provide new insights to physicians for dry eye diagnosis.