Abstract:
Due to their speed, deep learning-based deformable image registration methods are appealing. These methods fail to estimate larger displacements in complex deformation fields, requiring a multi-resolution strategy.
To solve this problem, we propose progressively training neural networks. We train smaller convolutional neural networks on lower-resolution images and deformation fields before training a large network on the registration task.
We add layers trained on higher-resolution data during training. We demonstrate that this training method allows a network to learn larger displacements without sacrificing registration accuracy and is less sensitive to large misregistrations than training the full network at once.
We apply random synthetic transformations to a training set of images to generate many ground truth example data and test the network on intrapatient lung CT registration.
We analyze the learned representations in the progressively growing network to determine how progressive learning affects training. Finally, progressive training improves registration accuracy for large and complex deformations.
Note: Please discuss with our team before submitting this abstract to the college. This Abstract or Synopsis varies based on student project requirements.
Did you like this final year project?
To download this project Code with thesis report and project training... Click Here
