Gene Expression Profile Prediction in Uveal Melanoma Using Deep Learning: a Pilot Study for Development of an Alternative Survival Prediction Tool.

We present a pilot study that utilized deep learning, leave-one-out cross validations and digitized cytology slides to predict survival and correlate gene expression profile of uveal melanomas, showing a point estimate of 75% accuracy on a patient-level analysis (15/20 patients, 95% CI 51 to 91).