DiffSR is a PyTorch-based framework for diffusion tensor imaging/dMRI super-resolution using spherical harmonic coefficient volumes. It supports training and inference l=2 single-shell SH representations of DTI data and is designed for research workflows involving diffusion preprocessing and tractography pipelines for heavily-degraded and/or low-field data.
Status: research code — APIs and configs may change.
- Super-resolution on SH coefficient volumes