Improved Spiral Projection MR Fingerprinting via Memory-Efficient Synergic Optimization of 3D Spiral Trajectory, Image Reconstruction, and Parameter Estimation (SOTIP)
This repository contains the official implementation of:
Jiaren Zou, Yun Jiang, Sydney Kaplan, Nicole Seiberlich, and Yue Cao.
Improved Spiral Projection MR Fingerprinting via Memory-Efficient Synergic Optimization of 3D Spiral Trajectory, Image Reconstruction, and Parameter Estimation (SOTIP).
IEEE Transactions on Medical Imaging, 2025.
[DOI: 10.1109/TMI.2025.3559467]
SOTIP is a memory- and computation-efficient model-based deep learning (MBDL) framework for full 3D spiral MR Fingerprinting (MRF).
It enables fast, high-resolution T1 and T2 mapping through:
- Memory-efficient MBDL reconstruction for non-Cartesian MRF.
- Joint optimization of temporal subspace image reconstruction and parameter estimation.
- Rotation angle optimization of 3D spiral sampling trajectories.
SOTIP-master/
fcnn.py # Fully Connected Neural Network for parameter estimation
network_experiments.sh # Bash script to organize training experiments
phantom_generation.py # Script to load phantoms and in vivo data
train_CNN.py # Main training script
env.yaml # Environment file. Additional requirement: MIRTorch (https://github.com/guanhuaw/MIRTorch)
unet/
model.py # U-Net for temporal subspace coefficient (TSC) image reconstruction
unet_parts.py # U-Net building blocks
utils/
data_processing.py # Data preprocessing utilities
data_analysis.py # Evaluation and analysis tools