Abstract:
The diagnosis and treatment of prostate diseases, especially prostate cancer, depend on accurate and reliable prostate gland segmentation using magnetic resonance (MR) imaging. Due to image appearance variability, imaging interference, and anisotropic spatial resolution, many automated segmentation methods, including deep learning, can improve segmentation performance.
This project proposes the 3D adversarial pyramid anisotropic convolutional deep neural network (3D APA-Net) for MR prostate segmentation. A generator (3D PA-Net) segments images, and a discriminator (6-layer convolutional neural network) distinguishes between segmentation results and ground truth.
The 3D PA-Net has a 3D ResNet encoder, an anisotropic convolutional decoder, and multi-level pyramid convolutional skip connections. Anisotropic convolutional blocks can exploit the 3D context information of MR images with anisotropic resolution, pyramid convolutional blocks address voxel classification and gland localization issues, and adversarial training regularizes 3D PA-Net to generate spatially consistent and continuous segmentation results.
We compared the proposed 3D APA-Net to several state-of-the-art deep learning-based segmentation methods on two public databases and a hybrid of the two. The proposed model outperforms the compared approaches on three databases and could be used in clinical workflows.
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
