Fundus blood vessel segmentation method based on branch attention and multi-model fusion

The invention relates to a CT image organ segmentation method based on convolutional neural network multi-dimensional fusion. The method comprises the steps of S1, training a Unet++ model through weights, training data and label data which are obtained through label calculation by means of an attention loss function; S2, training the Unet++ model by utilizing the training data, the label and a binary cross entropy loss function; S3, respectively obtaining two different segmentation results by utilizing the two obtained trained Unet++ models and the to-be-segmented data; and S4, fusing the twodifferent segmentation results. According to the method, the problem that some small blood vessels cannot be well segmented in the blood vessel segmentation problem of the fundus image is solved, so that the segmentation accuracy is improved.