Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
SDUM: A Scalable Deep Unrolled Model for Universal Cardiac MRI Reconstruction
Puyang Wang, Pengfei Guo, Keyi Chai, Jinyuan Zhou, Daguang Xu, Shanshan Jiang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use Restormer backbone instead of U-Net (ablation) | SDUM T=6 · CMRxRecon2025 Regular Task 1 | PSNR (dB): 28.734 -> 32.090 | Improved |
| use 2 down-sampling layers instead of 4 (ablation) | SDUM T=6 · CMRxRecon2025 Regular Task 1 | PSNR (dB): 31.544 -> 32.090 | Improved |
| use 2 down-sampling layers instead of 1 (ablation) | SDUM T=6 · CMRxRecon2025 Regular Task 1 | PSNR (dB): Training failed | Improved |
| widen single cascade from 7.2M to 78.7M params (ablation) | SDUM T=1 · CMRxRecon2025 Regular Task 1 | PSNR (dB): 27.605 -> 28.978 | Improved |
| widen single cascade beyond 42.2M params (ablation) | SDUM T=1 · CMRxRecon2025 Regular Task 1 | PSNR (dB): 28.728 -> 28.978 | No clear effect |
| use per-cascade CSME instead of single shared CSME (ablation) | SDUM · CMRxRecon2025 Regular Task 1 | PSNR (dB): 29.694 -> 30.205 | Improved |
| use SWDC instead of simple learnable DC (ablation) | SDUM T=6 · CMRxRecon2025 Regular Task 1 | PSNR (dB): 31.661 -> 32.090 | Improved |
| use SWDC instead of non-sampling-aware weighted DC (ablation) | SDUM T=6 · CMRxRecon2025 Regular Task 1 | PSNR (dB): 31.787 -> 32.090 | Improved |
| add universal conditioning (ablation) | SDUM T=1 Iter=3 · CMRxRecon2025 Regular Task 1 | PSNR (dB): 29.693 -> 30.071 | Improved |
| use 3 cascades instead of 3 iterations of one cascade (ablation) | SDUM · CMRxRecon2025 Regular Task 1 | PSNR (dB): 30.071 -> 30.205 | Improved |
| increase unrolled depth from T=1 to T=18 | SDUM · CMRxRecon2025 Regular Task 1 | PSNR (dB): 28.73 -> 33.18 | Improved |
| increase depth from T=10 to T=18 | SDUM · CMRxRecon2025 Regular Task 1 | PSNR (dB): 32.54 -> 33.18 | Improved |
Closest papers by meaning
- Fast MRI for All: Bridging Access Gaps by Training without Raw Data (NeurIPS 2025 · similarity 0.86)
- Efficient Noise Calculation in Deep Learning-based MRI Reconstructions (ICML 2025 · similarity 0.85)
- A Unified Model for Compressed Sensing MRI Across Undersampling Patterns (CVPR 2025 · similarity 0.85)
- Progressive Divide-and-Conquer via Subsampling Decomposition for Accelerated MRI (CVPR 2024 · similarity 0.85)
- Null Space Matters: Range-Null Decomposition for Consistent Multi-Contrast MRI Reconstruction (AAAI 2024 · similarity 0.84)
- Breaking Information Isolation: Accelerating MRI via Inter-sequence Mapping and Progressive Masking (AAAI 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
