Training deep‐learning segmentation models from severely limited data

Abstract

We demonstrated an effective data augmentation approach to train high-quality deep learning segmentation models from a limited number of well-contoured patient cases.