Contextual Augmentation Based on Metric-Guided Features for Ocular Axial Length Prediction

Ocular axial length (AL) measurement is important in ophthalmology because it should be considered prior to operations, such as strabismus surgery or cataract surgery, and the automation of AL with easily obtained retinal fundus images has been studied. However, performance deep learning methods inevitably depends on distribution data set used, lack an issue that needs addressed. In this study, we propose a framework for generating pairs their corresponding ALs improve inference. The generator’s encoder was trained independently using metric based information. A random vector zero padding were incorporated into generator increase creation flexibility, after which information inserted conditional We verified effectiveness by evaluating inference models training them combined comprising privately collected actual generated proposed method. Compared only set, mean absolute error standard deviation method decreased from 10.23 2.56 3.96 0.23, respectively, even smaller number layers prediction models.