Medical image segmentation method based on block chain and federal domain generalization

The invention relates to a medical image segmentation method based on block chain and federated domain generalization, and mainly solves the problems that a medical image segmentation model in the prior art is poor in generalization and lacks a credible execution environment and privacy protection. According to the implementation scheme, the method comprises the following steps: (1) downloading an initial network model from a block chain, performing training by using medical image data after interpolation is performed by using an amplitude spectrum of local data and amplitude spectrums transmitted by other federated domain data, and local data of the medical image data, and updating a local model; (2) adding differential privacy noise subjected to sensitivity calibration to the local model, uploading the local model to an interstellar file system (IPFS), and calculating hash by the system and taking the hash as a transaction; (3) carrying out transaction verification through a committee consisting of several