Research map / Detection and segmentation / CNN features
TopoFuse: Topology-Aware Tri-Planar Fusion for 3D Cryo-Electron Tomography Segmentation
Rohit Kumar Salla, Neelesh Gupta, Xingjian Li, Min Xu
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 differentiable topology projection (T=5) (ablation) | tri-planar + L_topo · EMPIAR-10499 ablation | BE0: 2.41 -> 1.11 | Improved |
| add differentiable topology projection (T=5) (ablation) | tri-planar + L_topo · EMPIAR-10499 ablation | Dice (%): 81.1 -> 85.2 | Improved |
| add Wasserstein topology loss without projection (ablation) | tri-planar · EMPIAR-10499 ablation | BE0: 2.88 -> 2.41 | Improved |
| add learned fusion without FiLM (ablation) | tri-planar · EMPIAR-10499 ablation | BE0: 3.10 -> 2.88 | Improved |
| add FiLM topology conditioning (ablation) | TopoFuse · EMPIAR-10499 ablation | BE0: 1.11 -> 1.10 | No clear effect |
| use learned topology prior instead of oracle at inference (ablation) | TopoFuse · EMPIAR-10499 ablation | BE0: 0.74 -> 1.09 | Hurt |
| increase projection steps from T=5 to T=10 (ablation) | TopoFuse · EMPIAR-10499 ablation | BE0: 1.11 -> 1.09 | No clear effect |
| reduce projection steps from T=5 to T=3 (ablation) | TopoFuse · EMPIAR-10499 ablation | BE0: 1.11 -> 1.21 | Hurt |
| TopoFuse vs BettiMatching soft loss | TopoFuse (learned) · three cryo-ET benchmarks (mean) | BE0: 2.68 -> 1.24 | Improved |
| TopoFuse vs BettiMatching soft loss | TopoFuse (learned) · three cryo-ET benchmarks (mean) | Dice (%): 80.3 -> 84.9 | Improved |
| non-differentiable PH post-processing instead of soft loss | tri-planar + TopoPost · EMD-0506 | Dice (%): 80.8 -> 79.0 Dice; BE0 2.29 -> 2.17 | Mixed |
| TopoFuse vs BettiMatching on synthetic noisy data | TopoFuse · SYN, SNR 0.10 | exact topology recovery (%): 61 -> 84 | Improved |
Closest papers by meaning
- Vox-UDA: Voxel-wise Unsupervised Domain Adaptation for Cryo-Electron Subtomogram Segmentation with Denoised Pseudo-Labeling (AAAI 2025 · similarity 0.88)
- At FullTilt: Real-Time Open-Set 3D Macromolecule Detection Directly from Tilted 2D Projections (NeurIPS 2026 · similarity 0.87)
- cryoSPHERE: Single-Particle HEterogeneous REconstruction from cryo EM (ICLR 2025 · similarity 0.87)
- CryoBench: Diverse and challenging datasets for the heterogeneity problem in cryo-EM (NeurIPS 2024 · similarity 0.86)
- CryoFM: A Flow-based Foundation Model for Cryo-EM Densities (ICLR 2025 · similarity 0.86)
- CryoSPIN: Improving Ab-Initio Cryo-EM Reconstruction with Semi-Amortized Pose Inference (NeurIPS 2024 · 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.
