MedCase-Structured
收藏资源简介:
MedCase-Structured是由System Inc.创建的一个临床现实合成数据集,旨在为电子健康记录(EHR)环境中的诊断推理提供结构化基准。该数据集基于MedCaseReasoning的约14,500个诊断案例,通过多阶段生成流程将非结构化文本转换为术语验证的HL7 FHIR R4患者捆绑包,最终包含1,408条有效数据,覆盖患者人口统计、症状、实验室结果等多维临床信息。数据创建过程结合了大型语言模型分阶段生成、术语接地验证与修复技术,确保FHIR资源的结构与语义一致性,成功转换率达82.5%。该数据集主要应用于临床决策支持系统的评估与训练,解决在结构化、可互操作的EHR格式下模型诊断准确性下降的问题,推动部署对齐的基准测试发展。
MedCase-Structured is a clinical real-world synthetic dataset developed by System Inc., which serves as a structured benchmark for diagnostic reasoning in electronic health record (EHR) environments. Built upon approximately 14,500 diagnostic cases sourced from MedCaseReasoning, it adopts a multi-stage generation pipeline to convert unstructured text into terminology-verified HL7 FHIR R4 patient bundles, ultimately yielding 1,408 valid entries covering multidimensional clinical information including patient demographics, symptoms, laboratory results, and other relevant clinical data. The dataset creation process integrates staged Large Language Model (LLM) generation, terminology grounding validation and correction technologies to guarantee the structural and semantic consistency of FHIR resources, with a successful conversion rate of 82.5%. This dataset is mainly applied to the evaluation and training of clinical decision support systems, addressing the problem of reduced diagnostic accuracy of models under structured and interoperable EHR formats, and advancing the development of deployment-aligned benchmark testing.
MedCase-Structured 数据集概述
基本信息
- 数据集名称: MedCase-Structured
- 存储位置: GitHub 仓库 SystemInternal/MedCase-Structured
- 相关论文: MedCase-Structured: A Text-to-FHIR Dataset for Benchmarking Diagnostic Reasoning in Clinically Realistic EHR Settings(arXiv: 2605.30295)
- 发布状态: 完整数据集将在 ICML 2026 Workshop on Structured Data for Health 举办前发布
数据集用途
- 用于在临床真实的电子健康记录(EHR)环境中评估诊断推理能力
- 提供 Text-to-FHIR 任务,即从文本到 FHIR(快速医疗互操作性资源)格式的转换基准
关联信息
- 论文预印本地址: https://arxiv.org/abs/2605.30295

- 1MedCase-Structured: A Text-to-FHIR Dataset for Benchmarking Diagnostic Reasoning in Clinically Realistic EHR SettingsSystem Inc. · 2026年



