Learning Deformable Registration of Medical Images with Anatomical Constraints

Repository of AC-RegNet, a new method to regularize CNN-based deformable image registration by considering global anatomical priors in the form of segmentation masks.

In this repository you can find the AC-RegNet source code and results produced for the paper "Learning deformable registration of medical images with anatomical constraints" (Neural Networks, 2020). You can download our paper from here.

Content

• CLI Application: An open source command line tool for chest x-ray image registration.
• AC-RegNet: Implementation of AC-RegNet model with TensorFlow.
• NIH Chest-XRay14 segmentations: Anatomical segmentation masks produced for NIH Chest-XRay14 dataset using AC-RegNet with a multi-atlas segmentation model.

Reference

If you use source code or results from this repository in your publication, please cite our paper:

• Mansilla, L., Milone, D. H., & Ferrante, E. (2020). Learning deformable registration of medical images with anatomical constraints. Neural Networks, 124, 269-279.

License

MIT