Method for detecting tree-shaped structure bifurcation key point in three-dimensional tomography image

The invention provides a method for detecting tree-shaped structure bifurcation key points in a three-dimensional tomography image, and relates to the field of medical image processing. According to the method, a three-dimensional tomography image containing an anatomical tree-shaped structure is acquired and preprocessed in an offline stage, the image is labeled, a corresponding prediction targetof bifurcation key point detection, tree-shaped structure segmentation and a branch vector field is generated according to a result, finally training data is formed and used for training a deep learning network, and a trained network is obtained; in the online stage, an image of the same type is acquired, the prediction results of a bifurcated key point heat map, the tree structure segmentation probability map and the branch vector field corresponding to the image are output by using the trained network, and the final key point detection position is calculated by applying the prediction results of the heat map and th