多模态双编码器检索用于自动化ICD编码
原标题：Multimodal Dual-Encoder Retrieval for Automated ICD Coding
AI 摘要
该研究提出一个两阶段框架用于自动化ICD编码：第一阶段用多模态双编码器检索模型，通过门控融合整合结构化EHR数据与非结构化临床笔记，将患者就诊记录与ICD编码元数据对齐到共享嵌入空间，并用FAISS近似最近邻搜索输出top-k候选编码；第二阶段用基于LLM的重排序器对候选编码重新打分并给出临床相关解释。在MIMIC-III等基准数据集上，该方法相比多模态双融合分类器基线在Micro-F1和Precision上均有提升，作者认为这是密集多标签分类的实用替代方案。
正文节选
Multimodal Dual-Encoder Retrieval for Automated ICD Coding Abstract Accurate International Classification of Diseases (ICD) coding is crucial for large-scale clinical research, documentation, and billing. There are three primary problems with current ICD prediction methods: (1) They are unable to comprehend multimodal patient data because they rely on either structured EHR data or unstructured clinical notes. (2) They also struggle with scalability to a larger amount of ICD codes (9K+ codes in I