发表机构
Northwestern University; Northwestern University Feinberg School of Medicine; Simpson Querrey Lung Institute for Translational Science（西北大学; 西北大学费恩伯格医学院; 辛普森奎雷转化科学肺研究所）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究提出分层多模态MoE模型，融合预训练影像专家与EHR，在ILD分类患者级交叉验证中AUC达0.8750±0.0443，优于REN和SwinUNETR，且提升了可解释性。
AI 中文摘要
混合专家（MoE）模型在学习到的路由机制下组合专用预测器，为利用医疗数据中的异质性提供了合理机制。我们提出一种用于间质性肺疾病（ILD）分类的分层多模态MoE，其通过两阶段门控机制将冻结的预训练影像专家与结构化电子健康记录（EHR）相融合。模态级门为影像和EHR预测分配患者特异性权重，而子门控模块将EHR分支分解为临床定义的特征组，并学习组特异性贡献。该设计保留了稳定的影像表示，同时支持输入依赖的临床加权和明确的EHR专业化。在严格的患者级交叉验证下，该模型在评估方法中取得最高平均AUC（0.8750±0.0443），仅影像的REN模型为0.8646，SwinUNETR为0.7685。该框架在解剖区域、影像-EHR利用及临床定义的EHR特征组上扩展了可解释性。
英文摘要
Mixture-of-experts (MoE) models combine specialized predictors under learned routing, offering a principled mechanism for leveraging heterogeneity in medical data. We present a hierarchical multimodal MoE for interstitial lung disease (ILD) classification that integrates a frozen, pre-trained imaging expert with structured electronic health records (EHR) via two-stage gating. A modality-level gate assigns patient-specific weights to imaging and EHR predictions, while a sub-gating module decomposes the EHR branch into clinically defined feature groups with learned, group-specific contributions. This design preserves stable imaging representations while enabling input-dependent clinical weighting and explicit EHR specialization. Under strict patient-level cross-validation, the model achieved the highest mean AUC among the evaluated methods (0.8750 +- 0.0443), compared with 0.8646 for imaging-only REN and 0.7685 for SwinUNETR. The framework extends interpretability across anatomical regions, imaging--EHR utilization, and clinically defined EHR feature groups.
Comments11 pages, 2 figures
Journal refMICCAI Machine Learning in Medical Imaging (MLMI 2026)
