Research map / Detection and segmentation / CNN features
Towards Cross-Modality Medical Image Segmentation with Online Mutual Knowledge Distillation
Kang Li, Lequan Yu, Shujun Wang, Pheng-Ann Heng
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 MRI as assistant modality with online mutual knowledge distillation and ensemble prediction | CT cardiac segmentation · MM-WHS 2017 | mean Dice: 0.8706 -> 0.9012 | Improved |
| Use the proposed method with ensemble prediction instead of Zhang et al. | CT cardiac segmentation · MM-WHS 2017 | mean Dice: 0.8850 -> 0.9012 | Improved |
| Use CT as assistant modality with online mutual knowledge distillation and ensemble prediction | MRI cardiac segmentation · MM-WHS 2017 | mean Dice: 0.8243 -> 0.8517 | Improved |
| Remove the Image Alignment Module (ablation) | Ours · MM-WHS 2017 CT target, MRI assistant | mean Dice (S_ens): 0.9012 -> 0.8829 | Hurt |
| Add Image Alignment Module (ablation) | Ours · MM-WHS 2017 CT target, MRI assistant | mean Dice (S_ens): 0.8829 -> 0.9012 | Improved |
| Remove synthetic-to-real knowledge distillation (ablation) | Ours · MM-WHS 2017 CT target, MRI assistant | mean Dice (S_ens): 0.9012 -> 0.8879 | Hurt |
| Add synthetic-to-real knowledge distillation (ablation) | Ours · MM-WHS 2017 CT target, MRI assistant | mean Dice (S_ens): 0.8879 -> 0.9012 | Improved |
| Remove real-to-synthetic knowledge distillation (ablation) | Ours · MM-WHS 2017 CT target, MRI assistant | mean Dice (S_ens): 0.9012 -> 0.7937 | Hurt |
| Add real-to-synthetic knowledge distillation (ablation) | Ours · MM-WHS 2017 CT target, MRI assistant | mean Dice (S_ens): 0.7937 -> 0.9012 | Improved |
| Increase assistant MRI training data from 5 to 20 volumes (ablation) | Ours · MM-WHS 2017 CT cardiac segmentation | mean Dice: 0.8916 -> 0.9012 | Improved |
| Increase assistant MRI training data from 10 to 20 volumes (ablation) | Ours · MM-WHS 2017 CT cardiac segmentation | mean Dice: 0.8945 -> 0.9012 | Improved |
| Ensemble the two mutually distilled segmentors rather than use S_real alone | Ours · MM-WHS 2017 CT target, MRI assistant | mean Dice: 0.8947 -> 0.9012 | Improved |
Closest papers by meaning
- Multi-modal Vision Pre-training for Medical Image Analysis (CVPR 2025 · similarity 0.87)
- Training Like a Medical Resident: Context-Prior Learning Toward Universal Medical Image Segmentation (CVPR 2024 · similarity 0.87)
- UniMRSeg: Unified Modality-Relax Segmentation via Hierarchical Self-Supervised Compensation (NeurIPS 2025 · similarity 0.87)
- Multi-view Evidential Learning-based Medical Image Segmentation (AAAI 2025 · similarity 0.87)
- Semantic-guided Masked Mutual Learning for Multi-modal Brain Tumor Segmentation with Arbitrary Missing Modalities (AAAI 2025 · similarity 0.87)
- Versatile Medical Image Segmentation Learned from Multi-Source Datasets via Model Self-Disambiguation (CVPR 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.
