A Comparison of Pre-trained Vision-and-Language Models for Multimodal Representation Learning across Medical Images and Reports | Synapse