Aiming at the problems of low image reading efficiency, crowded medical resources and fragmented cross-modal information of ophthalmic OCT and OCTA, a multimodal large model diagnostic framework for various diseases such as retinal diseases and optic nerve damage in glaucoma is constructed to fully explore the complementary information of various images such as SS-OCT structural images, SS-OCTA vascular images and optic nerve scans. Extract cross-modal joint features to improve the accuracy of automatic diagnosis.