An explainable biomedical foundation model via large-scale concept-enhanced vision–language pretraining
Nature
Read full postResearchers developed ConceptCLIP, an explainable biomedical foundation model trained on the large-scale MedConcept-23M dataset with 23 million image-text pairs. The model was evaluated across 78 diverse public, private, and newly curated biomedical datasets to advance vision-language understanding in healthcare.




