用向量对齐保留解剖结构关系，让小模型也能精准分割医学图像。
Displacement Preserving Relational Distillation for Robust Medical Segmentation
- 通过向量对齐保留教师模型的解剖轨迹方向与相对尺度
- 仅用5%参数和3%计算量，达85.46%骰子分数且边界误差更小
- 适合资源受限的临床环境部署高性能分割模型
准确的3D医学图像分割受限于解剖结构的多样性与高计算成本。尽管知识蒸馏（KD）可实现模型压缩，但传统方法常无法保留复杂结构，且易受背景噪声干扰。本文提出位移保持关系蒸馏（DPRD），通过向量对齐方式蒸馏潜在解剖轨迹，保留教师模型流形的方向与相对尺度，并在任务相关结构上锚定蒸馏过程，防止信号稀释。集成至nnU-Net后，DPRD在ISLES 2022与AMOS 2022基准测试中优于现有基线。尤其在AMOS数据集上，以85.46%的骰子分数超越高容量的MedNeXt教师模型，同时显著降低边界误差。尽管仅使用约5%的教师参数与约3%的浮点运算量，仍保持高结构一致性。该方法为资源受限的临床环境提供了鲁棒高效的高性能分割方案。代码已开源。
原文摘要 · Abstract (English)
Accurate 3D medical segmentation is limited by anatomical variability and high computational costs. While knowledge distillation (KD) offers a route for model compression, conventional methods often fail to preserve complex structures and are overwhelmed by background noise. We propose Displacement-Preserving Relational Distillation (DPRD), which distills latent anatomical trajectories via vector based alignment to preserve the orientation and relative scale of the teacher's manifold, and prevents signal dilution by anchoring distillation in task-relevant structures. Integrated into nnU-Net, DPRD outperforms established baselines on ISLES 2022 and AMOS 2022 benchmarks. Notably, on the AMOS dataset, DPRD achieves a Dice score of 85.46%, edging out the high-capacity MedNeXt teacher while significantly reducing boundary errors. Despite utilizing only ~5% of the teacher's parameters and ~3% of its FLOPs, our approach maintains high structural consistency. This provides a robust, efficient solution for deploying high performance segmenters in resource-constrained clinical environments. Code: https://github.com/ClinicaAlpha/DPRD-3D-MedSeg
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。