Research map / Vision-language and multimodal / CLIP and zero-shot
VoxTell: Free-Text Promptable Universal 3D Medical Image Segmentation
Maximilian Rokuss, Moritz Langenberg, Yannick Kirchhoff, Fabian Isensee, Benjamin Hamm, Constantin Ulrich, Sebastian Regnery, Lukas Bauer, Efthimios Katsigiannopulos, Tobias Norajitra, Klaus Maier-Hein
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 multi-stage fusion (3 decoder stages) instead of single-stage late fusion (ablation) | VoxTell · Validation split | Dice: 55.11 -> 60.16 | Improved |
| increase fusion stages from 3 to 5 (ablation) | VoxTell · Validation split | Dice: 60.16 -> 61.54 | Improved |
| add deep supervision across decoder stages (ablation) | VoxTell · Validation split | Dice: 61.54 -> 62.55 | Improved |
| scale training batch size from 2 to 128 (ablation) | VoxTell · Validation split | Dice: 62.55 -> 69.43 | Improved |
| use MaskFormer-style fusion (SAT) instead of Mask2Former (ablation) | single-stage fusion baselines · Validation split | Dice: 51.68 -> 55.11 | Improved |
| use VoxTell instead of SAT on report-derived prompts | VoxTell vs SAT · In-house SBRT lung tumor cohort | Dice: 0.0 -> 50.2 | Improved |
| use VoxTell multi-stage fusion instead of SAT single-stage after fine-tuning | VoxTell vs SAT · ReXGroundingCT validation | Dice: 13.1 -> 28.2 | Improved |
| use VoxTell instead of SAT on hit rate | VoxTell vs SAT · ReXGroundingCT validation | HIT5%: 49.8 -> 67.8 | Improved |
| Qwen3-Embedding-4B text encoder instead of SAT's original encoder (ablation) | SAT · Validation split | Dice | Improved |
Closest papers by meaning
- VOILA: Complexity-Aware Universal Segmentation of CT Images by Voxel Interacting with Language (AAAI 2025 · similarity 0.88)
- CAT: Coordinating Anatomical-Textual Prompts for Multi-Organ and Tumor Segmentation (NeurIPS 2024 · similarity 0.87)
- Enhancing vision-language models for medical imaging: bridging the 3D gap with innovative slice selection (NeurIPS 2024 · similarity 0.87)
- Better Tokens for Better 3D: Advancing Vision-Language Modeling in 3D Medical Imaging (NeurIPS 2025 · similarity 0.86)
- UniPro: Unified Multi-Mode Medical Image Segmentation from 2D Images to 3D Volumes via Propagation (NeurIPS 2026 · similarity 0.85)
- SegVol: Universal and Interactive Volumetric Medical Image Segmentation (NeurIPS 2024 · similarity 0.85)
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.
