How rfuentes89/flowmri works
FlowMRI-Net is a research implementation for reconstructing aortic and cerebrovascular 4D-flow MRI from undersampled measurements. The main workflow loads volunteer data, forms reconstruction inputs, runs a selectable learned model, and trains or tests it; a separate post-processing script performs background-phase correction and phase unwrapping. The repository also includes a CS-LLR reconstruction alternative and pretrained checkpoints. Edges reflect shown calls/imports or README statements; unsampled workflow details are not inferred.
Main entry points: main.py Train and test · postproc.py Post-processing · dataloader.py MRI dataset · CS_LLR_exec.py CS-LLR runner
Components 15
Reconstruction workflow
- Train and test
main.py - MRI dataset
utils/dataloader.py - SSDU partitioning
utils/partitioning.py - MRI operators
utils/misc_utils.py
Learned models
- FlowMRI-Net
networks/flowmri_net.py - FlowVN
networks/flowvn.py - Pretrained weights
weights
Alternative reconstruction
- CS-LLR runner
utils/CS_LLR_exec.py - CS-LLR solver
utils/CS_LLR_utils.py
Flow post-processing
- Post-processing
utils/postproc.py - Background correction
utils/execute_MSAC.py - Phase unwrapping
utils/unwrapFlow.py - Reconstruction output
- Corrected flow data
Other components
- Researcher
Connections 15
- Researcher → Train and test: configures
README.md - Train and test → MRI dataset: loads data
main.py - Train and test → FlowMRI-Net: selects model
main.py - Train and test → FlowVN: selects model
main.py - MRI dataset → SSDU partitioning: partitions k-space
utils/dataloader.py - MRI dataset → MRI operators: forms input images
utils/dataloader.py - Train and test → MRI operators: computes data consistency
main.py - Researcher → CS-LLR runner: configures
utils/CS_LLR_exec.py - CS-LLR runner → CS-LLR solver: uses solver functions
utils/CS_LLR_exec.py - Researcher → Post-processing: runs processing
- Post-processing → Reconstruction output: loads reconstruction
utils/postproc.py - Post-processing → Background correction: fits background phase
utils/postproc.py - Post-processing → Phase unwrapping: unwraps phase
utils/postproc.py - Post-processing → Corrected flow data: produces corrected data
utils/postproc.py - Researcher → Pretrained weights: selects checkpoint
README.md