用多教师对比蒸馏，让病理大模型变小变快，适合在边缘设备部署。
Multi-Teacher Contrastive Distillation for Edge-Efficient Pathology Foundation Models
- 从多个大模型提取嵌入特征，用对比学习方式蒸馏到轻量级学生模型。
- 小型模型在23项临床任务中接近原模型性能，参数量减少数倍，推理速度提升605倍。
- 适合医疗边缘设备部署，可在树莓派上实现高效病理图像分析。
计算病理学基础模型（PFMs）推动了全切片图像分析的发展，但其庞大体积和高推理成本限制了本地部署。本文提出MuCoDi框架，将多个冻结的病理模型（Virchow2、UNI2、H-Optimus-1）的瓦片嵌入特征，通过改进的MoCo v3对比蒸馏方法，压缩到轻量级移动模型（MobileOne、RepViT）中。在仅11.8K张全切片图像的1430万张瓦片上预训练，评估结果表明：基于RepViT的MuCoEdge-R2.3和R1.5分别达到71.0%外部AUROC，与最佳教师模型Virchow2（71.8%）相差仅0.8个百分点；其中MuCoEdge-R2.3在外部F1（51.8%）和AUPRC（53.3%）上表现优异。而仅含640万参数、1.12 GFLOPs的MuCoEdge-R1.0也实现了70.9%的AUROC。在树莓派5上，子百万参数的MobileOne模型单图推理速度比Virchow2快605倍，同时保持66.5%-66.9%的外部AUROC，证明了高质量病理表征可成功迁移到边缘设备。代码已公开。
原文摘要 · Abstract (English)
Computational pathology foundation models (PFMs) have advanced whole-slide image analysis. However, their size and inference cost hinder local deployment in pathology departments. We propose MuCoDi, a pretraining framework that distills frozen tile embeddings from multiple PFMs into compact edge-oriented encoders. Instead of regressing individual teacher features, MuCoDi trains lightweight MobileOne and RepViT students with a contrastive distillation objective adapted from MoCo v3, where cached Virchow2, UNI2, and H-Optimus-1 embeddings replace momentum-encoder keys. We pretrain students on 14.3M TCGA tiles from only 11.8K WSIs and evaluate frozen encoders on 23 clinically curated downstream classification tasks. RepViT-based MuCoEdge students retain near-teacher performance while reducing model size by orders of magnitude: MuCoEdge-R2.3 and MuCoEdge-R1.5 reach 71.0% external AUROC, within 0.8 percentage points of the best teacher (Virchow2, 71.8%), while MuCoEdge-R2.3 obtains the best external F1 and the second-best AUPRC (51.8% and 53.3%). MuCoEdge-R1.0 reaches 70.9% AUROC with only 6.4M parameters and 1.12 GFLOPs. On a Raspberry Pi 5, sub-million-parameter MobileOne students achieve up to 605-fold single-tile speedup over Virchow2 while retaining 66.5% to 66.9% external AUROC, demonstrating that PFM-quality pathology representations can be moved toward practical edge deployment. Code is available at https://anonymous.4open.science/r/mucodi-6243.
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。