Research map / Label-efficient and robust learning / Semi-supervised learning
Supervision by Denoising
Sean I. Young, Adrian Vasile Dalca, Enzo Ferrante, Polina Golland, Christopher A. Metzler, Bruce R. Fischl, Juan Eugenio Iglesias
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 SUD spatio-temporal denoising supervision to U-Net | nnU-Net cortical parcellation, 1 labeled image · HCP 810439 | Dice: 0.7423 -> 0.8206 | Improved |
| train under SUD instead of only post-denoising the baseline output (ablation) | nnU-Net · HCP 993675 parcellation | Dice: 0.793 -> 0.874 | Improved |
| add spatial denoising on top of stochastic ensembling (ablation) | nnU-Net · HCP 993675 parcellation | Dice: 0.805 -> 0.874 | Improved |
| add spatial denoising on top of stochastic ensembling (ablation) | nnU-Net · MMRR-3T7T-2-2 parcellation | Dice: 0.748 -> 0.800 | Improved |
| add spatial denoising on top of stochastic ensembling (ablation) | nnU-Net 3D brain · ABIDEII NYU | Dice: 0.773 -> 0.851 | Improved |
| add spatial denoising on top of stochastic ensembling (ablation) | nnU-Net 3D brain · COBRE 0040113 | Dice: 0.756 -> 0.826 | Improved |
| post-denoise baseline reconstructions instead of nothing (ablation) | nnU-Net · HCP 993675 parcellation | Dice: 0.7483 -> 0.793 | Improved |
| post-denoise baseline reconstructions (ablation) | nnU-Net · 3D brain COBRE 0040113 | Dice: 0.681 -> 0.708 | Improved |
| SUD over stochastic ensembling with 1 labeled image (ablation) | nnU-Net · Cortical parcellation | Dice and 95HD: improves mean Dice and Hausdorff by 0.04 and 18 pixels (~54%) | Improved |
| SUD over stochastic ensembling on labeled-set subject images (ablation) | nnU-Net 3D brain · Buckner39 vs ABIDE2 | Dice: ABIDE2, 5 points; Buckner39, 1 point | Mixed |
| SUD over stochastic ensembling (ablation) | nnU-Net 3D brain · Individual subject 95HD | 95HD: p = 10^-37 | Mixed |
| use 5 instead of 2 training label maps for denoiser (ablation) | Denoiser · 3D brain recon | Dice: 0.956 -> 0.969 | Improved |
Closest papers by meaning
- Denoising as Adaptation: Noise-Space Domain Adaptation for Image Restoration (ICLR 2025 · similarity 0.87)
- Self-Supervised Diffusion MRI Denoising via Iterative and Stable Refinement (ICLR 2025 · similarity 0.87)
- Score-based Self-supervised MRI Denoising (ICLR 2025 · similarity 0.87)
- Rethinking Reconstruction and Denoising in the Dark: New Perspective, General Architecture and Beyond (CVPR 2025 · similarity 0.87)
- DEFT: Efficient Fine-tuning of Diffusion Models by Learning the Generalised $h$-transform (NeurIPS 2024 · similarity 0.86)
- Efficient Noise Calculation in Deep Learning-based MRI Reconstructions (ICML 2025 · similarity 0.86)
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.
