Multi-mode MRI brain tumor image segmentation method

The invention discloses a multi-mode MRI brain tumor image segmentation method, and relates to the technical field of image segmentation. Comprising the following steps: acquiring a multi-modal brain tumor MRI image, and constructing an image data set; constructing a Unet network for image segmentation; inputting the data set into the Unet network, and training the Unet network by using the improved loss function; and inputting a brain tumor image to be segmented into the trained Unet network for segmentation. According to the method, the loss function is improved by introducing the difference between the overall gradient of the predicted tumor block and the overall gradient of the actual tumor block, and the network can be encouraged to generate a smoother segmentation boundary by minimizing the difference between the overall gradient of the predicted tumor block and the overall gradient of the actual tumor block. In addition, the overall gradient can provide information about tumor boundaries and forms, and