Pulmonary nodule benign and malignant prediction method based on three-dimensional deep learning network

The invention discloses a pulmonary nodule benign and malignant prediction method based on a three-dimensional deep learning network, and belongs to the field of image processing. The method comprisesthe following steps: preprocessing a lung CT image and generating multi-resolution input data; constructing a multi-resolution 3D dual-channel compression excitation deep learning network model; andperforming iteration by using a gradient descent method to obtain model parameters. According to the method, a deep learning network model combining a dual-path network and a compressive excitation network idea is adopted. The low-order features of the pulmonary nodule image can be repeatedly utilized to continuously generate new high-order combined features, the weights of the feature channels can be calibrated again, and the importance degree of different feature channels to network output is effectively described. The 3D multi-resolution data processing mode adopted by the invention can effectively solve the problem