Making the Most of Text Semantics to Improve Biomedical Vision--Language Processing