generating a plurality of 3-dimensional computed tomography angiography (3D CTA) reconstruction images of the patient; and

providing the plurality of 3D CTA reconstruction images of the patient to a prediction neural network configured to predict post-operative complications after endovascular aneurysm repair, wherein the prediction neural network is trained on a set of training data comprising a plurality of training examples, and wherein training the prediction neural network comprises:

generating one or more datasets including a plurality of post-operative 3D CTA reconstruction images of patients after endovascular aneurysm repair;

selecting a first set of images showing positive post-operative complications and a second set of images showing negative post-operative complications in accordance with a determined positive:negative ratio;

generating a set of downsampled images by downsampling the second set of images using a downsampling ratio of n;

generating augmented images by augmenting the first set of images and the set of downsampled images;

determining a training put from the augmented images; and

providing the training input to the prediction neural network.