There is none. But if the multimodal model is exposed to enough medical knowledge, it may be able to interpret images without specific training