aboutill/sparc repository overview

⁠SPARC

Automated, end-to-end pipeline for reconstructing 3D+time fetal cardiac cine MRI from Doppler-gated 2D stacks, combining physics-informed slice-to-volume reconstruction with deep learning (segmentation and reorientation).

Research software accompanying a manuscript: https://doi.org/10.48550/arXiv.2608.18616⁠. Full documentation, source, and reproducibility materials: https://github.com/aboutill/SPARC⁠

⁠Quick start

This image is designed to be run through the SPARC CLI wrapper: it handles GPU passthrough, volume mounts, and (where needed) GUI forwarding, which a bare docker run does not do for you. Install from the GitHub repo:

git clone https://github.com/aboutill/SPARC
cd SPARC
make install
SPARC pipeline --input /path/to/dicom --output /path/to/results --mode semi_auto

If you just want to pull the image directly:

docker pull aboutill/sparc:v1.0.0

⁠What's included

• Python 3.11, PyTorch + MONAI (segmentation and reorientation networks)
• MRtrix3 3.0.8 (denoising, Gibbs-ringing correction)
• ITK-SNAP 3.8.0 (interactive review)
• SVR-lite 1.1.0 (slice-to-volume reconstruction), maintained as a separate project: https://github.com/baby-MedIA/svr-lite⁠

⁠Requirements

• Docker with the NVIDIA Container Toolkit, for GPU acceleration (strongly recommended, the pipeline runs on CPU, but substantially slower)

⁠Tags

• v1.0.0 matching the manuscript's reported version.