从注入到同化的蒸馏用于医学图像分割
From Infusion to Assimilation Distillation for Medical Image Segmentation
把知识蒸馏拆成「注入+同化」两阶段：先用类加权原型对齐软标签蒸馏对齐师生预测分布，再用对比式语义自优化对编码器特征施加反向约束，可作为在算力受限场景下压缩 SAM 类基础模型到轻量分割网络的通用蒸馏范式。
原文依据 soft-label distillation is combined with class-weighted prototype alignment strategy
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论尽管基础模型（如 SAM）在医学图像分割中表现优异，但其高计算复杂度限制了部署。知识蒸馏（KD）使轻量模型能够继承大模型的表征能力，从而缓解这一问题。现有 KD 方法提升了学生模型性能，但由于教师-学生特征优势不同，这些方法忽视了在知识迁移后对学生语义信息的自适应内化与整合，导致知识同化不佳，限制了性能增益与泛化能力。为解决这一局限，我们提出了一种新的医学图像分割框架，即从注入到同化的蒸馏（IAD）。在知识注入阶段（KIS）中，为在语义上对齐教师-学生的预测分布，将软标签蒸馏与类加权原型对齐策略相结合。在知识同化阶段（KAS）中，为促进自适应语义同化，一种对比式语义自优化策略通过正负样本对精炼学生预测，并对编码器特征施加反向约束以增强语义一致性。IAD 在 Synapse 上取得 4.32%、ACDC 上 1.85%、Polyp 数据集上 2.42% 的 DICE 提升，并在 ISIC2018、PH2、BUSI 和 STU 数据集上带来平均 4.16% 的泛化增益，优于主流 KD 方法。代码见 https://github.com/hjklearn/IAD。
英文原文摘要 1464 字符
Although foundation models (e.g. SAM) perform remarkably in medical image segmentation, its high computational complexity limits deployment. Knowledge distillation (KD) allows lightweight models to inherit the representational capabilities of large models, thereby mitigating this issue. Existing KD methods enhance student performance, but due to teacher-student different feature advantages, they neglect to internalize and integrate student's semantic information adaptively after knowledge transfer, causing poor knowledge assimilation and limiting gains and generalization. To address this limitation, we propose a novel medical image segmentation framework, which is infusion to assimilation distillation (IAD). In Knowledge Infusion Stage (KIS), to semantically align teacher-student prediction distributions, soft-label distillation is combined with class-weighted prototype alignment strategy. In Knowledge Assimilation Stage (KAS), to promote adaptively semantic assimilation, a contrastive semantic self-optimization strategy refines student predictions through positive and negative sample pairs and imposes reverse constraints on encoder features to enhance semantic consistency. IAD achieves DICE gains of 4.32% on Synapse, 1.85% on ACDC, and 2.42% on Polyp datasets, and delivers an average 4.16% generalization gain on ISIC2018, PH2, BUSI, and STU datasets, outperforming mainstream KD methods. Code is available at https://github.com/hjklearn/IAD.