发表机构
New York University; University of Chicago（纽约大学; 芝加哥大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对临床知识图谱构建难题，提出结合多智能体关系提议等技术的Clinical Graph-JEPA框架，在MIMIC-IV数据集上验证其提升了留一边缘恢复的MRR。
AI 中文摘要
临床记录包含关于患者状态的丰富证据，但将这些证据转化为可靠的结构化知识图谱仍然困难，因为提取错误、本体不匹配、关系缺失和时间歧义会传播到下游系统。我们提出一种临床知识图谱构建与优化框架，结合多智能体关系提议、本体感知归一化、确定性证据评分以及基于JEPA的潜在优化。我们不将临床知识图谱视为静态提取产物，而是将其视为预测性患者状态表示。对于每次入院，系统从结构化的MIMIC-IV记录和推断的临床交叉链接构建经证据评分的图谱，然后学习从观测到的图谱上下文恢复预留的临床关系。我们采用无泄露的留一边缘恢复（MRR和Hits@k）以及预留批次掩码评估（AUC和MRR）来评估优化器。为分离出院记录上下文的贡献，我们将无记录嵌入的配置与仅向基于记录的实体注入真实出院记录表示的记录增强配置进行比较。在相同队列和评估协议下，基于实体的记录注入使整体留一MRR提升了31%的相对改进。
英文摘要
Clinical records contain rich evidence about patient state, but converting that evidence into reliable, structured knowledge graphs remains difficult because extraction errors, ontology mismatch, missing relations, and temporal ambiguity can propagate into downstream systems. We propose a clinical knowledge graph construction and refinement framework that combines multi-agent relation proposal, ontology-aware normalization, deterministic evidence scoring, and JEPA-based latent refinement. Rather than treating a clinical knowledge graph as a static extraction artifact, we treat it as a predictive patient-state representation. For each admission, the system constructs an evidence-scored graph from structured MIMIC-IV records and inferred clinical cross-links, then learns to recover held-out clinical relations from the observed graph context. We evaluate the refiner with leakage-free leave-one-out edge recovery (MRR and Hits@k) and held-out batch-mask evaluation (AUC and MRR). To isolate the contribution of discharge-note context, we compare a note-embedding-free configuration with a note-augmented configuration that injects real discharge-note representations only into note-grounded entities. Under the same cohort and evaluation protocol, entity-grounded note injection improves overall leave-one-out MRR by 31% relative improvement.
CommentsAccepted at WM@Booth 2026
