Research map / Image generation / Image restoration
A Theoretically Guaranteed Deep Optimization Framework for Robust Compressive Sensing MRI
Risheng Liu, Yuxi Zhang, Shichao Cheng, Xin Fan, Zhongxuan Luo
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 |
|---|---|---|---|
| Add the optimal-condition checking module to the fidelity, data-driven, and prior modules (ablation) | Deep CS-MRI framework · T1-weighted data, Radial sampling 20% | reconstruction accuracy | Improved |
| Combine the fidelity and data-driven modules (ablation) | Deep CS-MRI framework · T1-weighted data, Radial sampling 20% | early-iteration loss | Improved |
| Add the manual prior module after fidelity and data-driven modules (ablation) | F→N CS-MRI variant · T1-weighted data, Radial sampling 20% | reconstruction accuracy | Improved |
| Use the three modules without optimal-condition checking (ablation) | F→N→P CS-MRI variant · T1-weighted data, Radial sampling 20% | convergence | Hurt |
| Replace the two-stage Rician-specific denoiser with direct noisy-to-clean difference learning (ablation) | Rician denoising network · Real-world CS-MRI reconstruction | PSNR: 28.71 -> 17.44 | Hurt |
| Replace the two-stage Rician-specific denoiser with separate real/imaginary Gaussian-noise estimation (ablation) | Rician denoising network · Real-world CS-MRI reconstruction | PSNR: 28.71 -> 19.25 | Hurt |
| Treat Rician noise as Gaussian rather than model its structure (ablation) | Rician denoising network · Real-world CS-MRI reconstruction | PSNR: 28.71 -> 24.41 | Hurt |
| Use the two-stage Rician-specific learnable architecture instead of RiceOptVST | MRI denoiser · T1-weighted MRI denoising | PSNR: 30.34 -> 31.72 | Improved |
| Use the proposed framework rather than the strongest listed baseline | CS-MRI reconstruction · Chest, Cartesian mask 30% | PSNR: 27.73 -> 28.22 | Improved |
| Use the proposed framework rather than the strongest listed baseline | CS-MRI reconstruction · Cardiac, Cartesian mask 30% | PSNR: 33.84 -> 35.69 | Improved |
| Use the proposed framework rather than the strongest listed baseline | CS-MRI reconstruction · Renal, Cartesian mask 30% | PSNR: 33.74 -> 34.79 | Improved |
| Use the proposed framework rather than BM3D-MRI followed by the same learned denoiser | Rician-noisy CS-MRI reconstruction · T1-weighted | PSNR: 27.64 -> 28.05 | Improved |
Closest papers by meaning
- Information-Theoretic Optimization for Task-Adapted Compressed Sensing Magnetic Resonance Imaging (TPAMI 2026 · similarity 0.91)
- MRI Reconstruction with Interpretable Pixel-Wise Operations Using Reinforcement Learning (AAAI 2020 · similarity 0.91)
- A Unified Model for Compressed Sensing MRI Across Undersampling Patterns (CVPR 2025 · similarity 0.88)
- Efficient Noise Calculation in Deep Learning-based MRI Reconstructions (ICML 2025 · similarity 0.87)
- Fast MRI for All: Bridging Access Gaps by Training without Raw Data (NeurIPS 2025 · similarity 0.86)
- UFC-Net: Unrolling Fixed-point Continuous Network for Deep Compressive Sensing (CVPR 2024 · similarity 0.85)
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.
