The invention belongs to the technical field of aortic aneurysm auxiliary diagnosis, and particularly relates to a deep learning-based computed tomography aortic aneurysm auxiliary diagnosis method, which comprises the following steps of: 1, data acquisition: firstly, retrospectively collecting a plurality of common CT scans with aorta from a hospital area of a hospital; 2, data processing: dividing the data into three data sets: a training set, an internal test set and an external test set; and 3, establishing a model, constructing an auxiliary diagnosis tool by using an Attention-Unet convolutional neural network, evaluating the risk, detection sensitivity, specificity and accuracy of the aortic aneurysm in a test set by using the auxiliary diagnosis tool. The method is reasonable in design and has good diagnosis capability on aortic aneurysm; when the method is used in combination with radiologists, the film reading and decision-making capabilities of the radiologists can be remarkably enhanced, and the overall strength of imaging departments is improved. Therefore, the performance of the advanced technology proves that the non-invasive, cheap and convenient method has the potential of clinical tests.