Dealing with Small Datasets for Deep Learning in Medical Imaging: An Evaluation of Self-Supervised Pre-Training on CT Scans Comparing Contrastive and Masked Autoencoder Methods for Convolutional Models.