Research map / Vision-language and multimodal / Cross-modal learning
XPos3R: Cross-Modal Transformer for Intraoperative 2D/3D Registration
Shiyan Su, Ruyi Zha, HONGDONG LI, Xuelian Cheng, Zongyuan Ge
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 |
|---|---|---|---|
| use RayDINO instead of DINOv2 as image encoder (ablation) | XPos3R · DeepFluoro (feedforward) | TRE (mm): 29.77 -> 24.16 | Improved |
| use pretrained RayDINO instead of training image encoder from scratch (ablation) | XPos3R · DeepFluoro (feedforward) | TRE (mm): 36.72 -> 24.16 | Improved |
| use lightweight 3D CNN trained from scratch instead of SuPreM 3D encoder (ablation) | XPos3R · DeepFluoro (feedforward) | TRE (mm): 74.03 -> 24.16 | Improved |
| use 3D CNN instead of SuPreM (ablation) | XPos3R · DeepFluoro (feedforward) | RE (deg): 2.607 -> 1.352 | Improved |
| add global patient embedding (ablation) | XPos3R · DeepFluoro (feedforward) | TRE (mm): 32.58 -> 24.16 | Improved |
| add global patient embedding (ablation) | XPos3R · DeepFluoro (feedforward) | RE (deg): 2.012 -> 1.352 | Improved |
| patient-specific finetuning of pretrained XPos3R (ablation) | XPos3R · DeepFluoro (feedforward) | TRE (mm): 24.16 -> 10.56 | Improved |
| patient-specific finetuning before test-time optimization (ablation) | XPos3R + TTO · DeepFluoro (feedforward + TTO) | TRE (mm): 3.288 vs. 3.275 mm | No clear effect |
| add XPos3R feedforward initialization to DiffDRR optimization | DiffDRR · DeepFluoro (TTO) | TRE (mm): 200.2 -> 3.288 | Improved |
| use XPos3R pretrained instead of XVR pretrained in feedforward | XPos3R · DeepFluoro (feedforward) | RE (deg): reducing RE and TE by 5.39 and 66.59 mm | Improved |
| use XPos3R pretrained instead of XVR finetuned for 2D fiducial error | XPos3R · DeepFluoro (feedforward) | PE (mm): 3.860 -> 7.412 | Hurt |
| add TTO to XPos3R | XPos3R · DeepFluoro (feedforward) | TRE (mm): 24.16 -> 3.288 | Improved |
Closest papers by meaning
- Intraoperative 2D/3D Image Registration via Differentiable X-ray Rendering (CVPR 2024 · similarity 0.89)
- PolyPose: Deformable 2D/3D Registration via Polyrigid Transformations (NeurIPS 2025 · similarity 0.87)
- JSSR: A Joint Synthesis, Segmentation, and Registration System for 3D Multi-Modal Image Alignment of Large-scale Pathological CT Scans (ECCV 2020 · similarity 0.85)
- LiFT: Lifted Inter-slice Feature Trajectories for 3D Image Generation from 2D Generators (NeurIPS 2026 · similarity 0.83)
- Better Tokens for Better 3D: Advancing Vision-Language Modeling in 3D Medical Imaging (NeurIPS 2025 · similarity 0.83)
- CARL: A Framework for Equivariant Image Registration (CVPR 2025 · similarity 0.82)
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.
