UEPS: Robust and Efficient MRI Reconstruction (Pre-trained Model and Demo Data)

This record contains the pre-trained weights (ckpt_pick.pth), sample MRI data, and visual results for the UEPS framework.

UEPS is a novel deep unrolled model (DUM) architecture designed for robust and efficient MRI reconstruction. It features three key innovations: (i) an Unrolled Expanded (UE) design that eliminates coil sensitivity maps (CSM) dependency by expanding multi-coil data to the batch dimension; (ii) progressive resolution, which leverages k-space-to-image mapping for efficient coarse-to-fine refinement; and (iii) sparse attention tailored to MRI's 1D undersampling nature.

Files included:
* ckpt_pick.pth: Pre-trained model weights.
* demo_data.zip: Sample MRI data for quick testing and visualization.
* reconstruction_examples.zip: Qualitative visualization examples of reconstructed slices. The image filenames follow the format '{slice_index}_nmse_{value}_psnr_{value}_ssim_{value}.png' (e.g., 140_nmse_0.00663_psnr_42.3608_ssim_0.9784.png).

For the official PyTorch implementation, please visit our GitHub repository: https://github.com/HongShangGroup/UEPS