Research map / Label-efficient and robust learning / Self-supervised and contrastive
MedGMAE: Gaussian Masked Autoencoders for Medical Volumetric Representation Learning
Xueming Fu, Fenghe Tang, Rongsheng Wang, Yingtai Li, Lixia Han, Jian Lu, Zihang Jiang, S Kevin Zhou
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 voxel-based SSL pre-training (ablation) | UNETR · AMOS | DSC (%): 77.02 -> 83.61 | Improved |
| add Gaussian-based SSL on top of voxel SSL (ablation) | UNETR · AMOS | DSC (%): 83.61 -> 84.90 | Improved |
| add Gaussian-based SSL on top of voxel SSL (ablation) | UNETR · FLARE'22 | DSC (%): 82.56 -> 83.77 | Improved |
| add Gaussian-based SSL on top of voxel SSL (ablation) | UNETR · SegTHOR | DSC (%): 88.52 -> 89.15 | Improved |
| raise mask ratio from 25% to 75% (ablation) | MedGMAE · AMOS 100% | DSC (%): 80.03 -> 84.90 | Improved |
| raise mask ratio from 75% to 85% (ablation) | MedGMAE · AMOS 100% | DSC (%): 84.90 -> 85.17 | No clear effect |
| use 512 Gaussians instead of 256 (ablation) | MedGMAE · AMOS 100% | DSC (%): 79.99 -> 84.90 | Improved |
| use 512 Gaussians instead of 1024 (ablation) | MedGMAE · AMOS 100% | DSC (%): 78.87 vs 84.90 | Hurt |
| use mask ratio 75% instead of 25% for zero-shot init (ablation) | 3DGR reconstruction · AAPM-Mayo 120 proj | training time (min): 541.5 -> 357.0 | Improved |
| use K=512 instead of K=1024 for zero-shot init (ablation) | 3DGR reconstruction · AAPM-Mayo 120 proj | training time (min): 546.7 -> 357.0 | Improved |
| initialize 3DGR with MedGMAE Gaussians | 3DGR · AAPM-Mayo 80 projections | training time (min): 397 -> 303 | Improved |
| initialize 3DGR with MedGMAE Gaussians | 3DGR · AAPM-Mayo 80 projections | PSNR (full): 44.6 -> 43.9 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- VoCo — VoCo: A Simple-yet-Effective Volume Contrastive Learning Framework for 3D Medical Image Analysis (CVPR 2024) · better on 8 of 9 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Learning General-purpose Biomedical Volume Representations using Randomized Synthesis (ICLR 2025 · similarity 0.85)
- Raptor: Scalable Train-Free Embeddings for 3D Medical Volumes Leveraging Pretrained 2D Foundation Models (ICML 2025 · similarity 0.84)
- Revisiting MAE Pre-training for 3D Medical Image Segmentation (CVPR 2025 · similarity 0.83)
- SSC-VAE: Structured Sparse Coding Based Variational Autoencoder for Detail Preserved Image Reconstruction (AAAI 2025 · similarity 0.83)
- Why Invariance is Not Enough for Biomedical Domain Generalization and How to Fix It (NeurIPS 2026 · similarity 0.83)
- Better Tokens for Better 3D: Advancing Vision-Language Modeling in 3D Medical Imaging (NeurIPS 2025 · similarity 0.83)
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.
