Large-Scale Domain-Specific Pretraining for Biomedical Vision-Language Processing