AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

Spatial-SAM: Spatially Consistent 3D Electron Microscopy Segmentation with SDF Memory and Semi-Supervised Learning

Yikai Huang, Renmin Han, Yuxuan Wang, Youcheng Cai, Ligang Liu

CVPR 2026 · Label-efficient and robust learning › Semi-supervised learning · approach: Semi-supervised medical segmentation

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
replace SAM2 logit memory with SDF memory (unidirectional) (ablation)Spatial-SAM · MitoEM-R (1/64 labels)Dice: 92.62 -> 93.62Improved
make SDF memory bidirectional instead of unidirectional (ablation)Spatial-SAM · MitoEM-R (1/64 labels)Dice: 93.62 -> 94.45Improved
exclude the current slice's own SDF memory from the neighborhood (ablation)Spatial-SAM · MitoEM-R (1/64 labels)Dice: 94.11 -> 94.45Improved
increase labeled slices from 1/64 to 1/4 (ablation)Spatial-SAM · MitoEM-RDice: 94.45 -> 95.35Improved
use Spatial-SAM with 1/64 labels instead of CPS U-Netsemi-supervised 3D EM segmentation · four-dataset averageDice: +8.07% Dice, +11.49% mIoUImproved
Spatial-SAM with 1/64 labels versus fully supervised SAM4EMSpatial-SAM · four-dataset averageDice: -0.09% Dice, -0.10% mIoUNo clear effect
more dual-track iterations (SDF and mask training) beyond three (ablation)Spatial-SAM · validation mIoUmIoUNo clear effect
SAM2-initialized pseudo-labels with dual-track refinementSpatial-SAM · OOMLN (nuclei)Dice: 92.91 -> 98.14 (vs 3D U-Net full labels)Improved

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