Two pathways are utilized to generate simulated data sets for model training: forward projection by DeepDRR, and 2D merging with X-ray images.
The link for downloading pretrained models is as follows: https://faubox.rrze.uni-erlangen.de/getlink/fiHnt9NBbGcfboiciJt1Z1/
Run test.py to test models on one simulated case.
Paper for this project:
Fan, Fuxin, et al. "Simulation-Driven Training of Vision Transformers Enables Metal Artifact Reduction of Highly Truncated CBCT Scans.“ (Submitted to Medical Physics)