Weak supervision medical image segmentation registration cooperation method based on comparative learning and spatial coding

The invention provides a weak supervision medical image segmentation registration cooperation method based on comparative learning and spatial coding, and the method comprises the steps: firstly constructing an image segmentation model, and pre-training the segmentation model through a small number of annotated images; acquiring medical image data and organ segmentation labels thereof, and dividing the medical image data and the organ segmentation labels into floating images and fixed images; predicting an image segmentation label without the segmentation label by using the segmentation model; calculating a spatial code of the image by using the predicted segmentation label for training a registration network; corresponding information of a segmentation labeling calculation structure is used for training a pixel-level Siamese network; using the registration model to generate augmented data, combining the augmented data with the Siamese network to perform retraining of the segmentation model, and repeating tra