AI Paper Atlas

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

ICLR 2026 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Self-supervised neuroimaging representations

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 changedWhereMeasuredEffect
add voxel-based SSL pre-training (ablation)UNETR · AMOSDSC (%): 77.02 -> 83.61Improved
add Gaussian-based SSL on top of voxel SSL (ablation)UNETR · AMOSDSC (%): 83.61 -> 84.90Improved
add Gaussian-based SSL on top of voxel SSL (ablation)UNETR · FLARE'22DSC (%): 82.56 -> 83.77Improved
add Gaussian-based SSL on top of voxel SSL (ablation)UNETR · SegTHORDSC (%): 88.52 -> 89.15Improved
raise mask ratio from 25% to 75% (ablation)MedGMAE · AMOS 100%DSC (%): 80.03 -> 84.90Improved
raise mask ratio from 75% to 85% (ablation)MedGMAE · AMOS 100%DSC (%): 84.90 -> 85.17No clear effect
use 512 Gaussians instead of 256 (ablation)MedGMAE · AMOS 100%DSC (%): 79.99 -> 84.90Improved
use 512 Gaussians instead of 1024 (ablation)MedGMAE · AMOS 100%DSC (%): 78.87 vs 84.90Hurt
use mask ratio 75% instead of 25% for zero-shot init (ablation)3DGR reconstruction · AAPM-Mayo 120 projtraining time (min): 541.5 -> 357.0Improved
use K=512 instead of K=1024 for zero-shot init (ablation)3DGR reconstruction · AAPM-Mayo 120 projtraining time (min): 546.7 -> 357.0Improved
initialize 3DGR with MedGMAE Gaussians3DGR · AAPM-Mayo 80 projectionstraining time (min): 397 -> 303Improved
initialize 3DGR with MedGMAE Gaussians3DGR · AAPM-Mayo 80 projectionsPSNR (full): 44.6 -> 43.9No 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.

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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy