Medical image segmentation method based on multiple tasks

The invention discloses a medical image segmentation method based on multiple tasks, and the method comprises the steps: enabling a transform to be integrated into YOLOv5, and constructing a transYL network; performing multi-task segmentation of shadow region detection and boundary detection by adopting the constructed transYL network; the method comprises the following steps of: establishing a Mean Teaching network transYLmtMT, wherein a Student network and a Teaching network are both the established transYL network; the transYL network is trained, and finally trained model parameters are stored; the trained Student network is tested, and values of corresponding test indexes are recorded; and the hyper-parameters are properly adjusted according to a test result, the training steps are carried out again until a satisfactory test result exists, and then the tested transYLmtMT network is used for medical image segmentation. According to the method, the segmentation result of the medical image infection area is