Age-related ocular surface modifications assessment combining thermal infrared and deep learning approach.

Age-related diseases such as glaucoma, diabetic retinopathy, and macular degeneration remain the leading causes of low, vision in developed countries. Early detection of such diseases can prevent the risk of progression to blindness. To this end, regular check-ups are encouraged to favor timely eye disease diagnosis. Yet, conducting routine large-scale eye screening can be difficult and time-consuming. In this study, a novel, fast and automatic approach for age-related ocular surface modifications (AR-OSM) assessment is proposed. Indeed, accurate AR-OSM detection in the healthy population may allow to establish age-matched normal ranges, valuable for the preliminary identification of age-related diseases. The task was performed combining thermal infrared (IR) imaging of the eye with artificial intelligence techniques. Thermal IR imaging enables non-invasive real-time imaging of the ocular surface temperature (OST). OST is influenced by ocular factors like the tear film, blood flow perfusion, heat conduction, and convection of the aqueous humor, thus providing significant information on eye health. Ninety-two healthy subjects participated in the experiment (age: 20-90 years-old). A Deep convolutional neural network (DCNN) model was implemented to predict the subjects' age based on their eye IR-image. The DCNN was able to predict the participants' age with a good level of accuracy, reporting a correlation between real and predicted age of r=0.82 and RMSE=9.9years. In conclusion, this method allows an accurate AR-OSM evaluation usable for early recognition of eyes at risk for age-related disease.