通过自适应融合提升病理大模型的准确率与可解释性
Understanding Synergistic Interactions among Pathology Foundation Models via Adaptive Fusion
- 用低维压缩和样本感知门控机制融合多个冻结的病理大模型
- 在三个公开数据集上均超越单个模型和传统融合方法
- 可解释性可视化揭示模型偏好与组织表型的协同关系
病理基础模型（PFMs）通过大规模病理图像自监督预训练，获得强大的切片级表征。然而，不同PFM在数据、架构和目标选择上差异显著且常不透明，导致潜在表征偏差，限制鲁棒性并模糊各模型专长。本文提出AdaFusion，一种轻量级自适应融合框架，通过（1）低维特征压缩和（2）样本条件门控模块，动态重加权模型间（及通道间）贡献。不仅提升预测准确率，还提供基于贡献的可解释性分析，揭示模型特异性偏好与跨组织表型的协同作用。在涵盖治疗反应预测、前列腺癌分级和空间基因表达推断的三个公开基准上验证，AdaFusion持续优于单一PFM及其他融合基线，并生成与形态模式一致的可解释组织可视化。代码已开源：https://github.com/xyx-98/PathoOracle。
原文摘要 · Abstract (English)
Pathology foundation models (PFMs) provide strong tile-level representations via self-supervised pre-training on large-scale pathology images. Yet, PFMs are developed under diverse and often opaque data, architecture, and objective choices, inducing latent representational biases that limit robustness and obscure what each model specialises in. We present AdaFusion, a lightweight adaptive fusion framework that integrates complementary signals from multiple frozen PFMs through (1) low-dimensional feature compression and (2) a sample-conditioned gating module that reweights model-wise (and optionally channel-wise) contributions. Beyond improving predictive accuracy, AdaFusion provides contribution-driven interpretation that offers evidence consistent with model-specific preferences and synergistic interactions across tissue phenotypes. We evaluate AdaFusion on three public benchmarks spanning treatment response prediction, prostate cancer grading, and spatial gene expression inference. AdaFusion consistently outperforms individual PFMs and other fusion baselines, while providing interpretable tissue visualisation which aligns model preferences with morphological patterns. Code is available at: https://github.com/xyx-98/PathoOracle.
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。