Method and device for constructing medical image large model based on federated learning

The invention discloses a method and device for constructing a medical image large model based on federated learning, and belongs to the technical field of medical artificial intelligence. The method is applied to a server and comprises the steps of performing pre-training by utilizing medical image public data to obtain a global initialization large model; distributing the globally initialized large model to each client; and generating a global large model by using the respective corresponding local large model uploaded by each client, the local large model being obtained by performing optimization training on the global initialization large model by each client using the local medical image data. Through a federal learning mode, the collaborative value of centralized public data and private data of all hospitals is fully played, efficient and safe co-construction of a medical image large model is realized, the stability and generalization ability of the model are improved, the development of artificial inte