This project aims at implementing a deep learning framework for 3D reconstruction of an anatomy of interest given 2D X-ray images as input. During the project the following tasks should be implemented:
- Compile a list of publicly available datasets of 2D x-ray images with ground truth reconstructions preferably from the Heart Vessels or Heart Anatomy.
- Generate a large amount of training images in the form of DRRs from CT/MRI data
- Implement a Deep Learning network that fulfills the tasks of 3D reconstruction from 2D x-ray images such as [1,2].
- Validate your results in comparison to the ground truth and present your results.
[1] A. Vlontzos, et al. “3D Probabilistic Segmentation and Volumetry from 2D projection images.” International Workshop on Thoracic Image Analysis, 2020