DeidSharp 开源工具包：本地 26B MoE LLM 加规则检测器做临床文本去标识化
A local language model layered with rule-based and neural detectors for clinical text de-identification: an open-source toolkit and evaluation
做了个开源的 .NET 临床去标识化工具包，规则加 NER 加本地 26B LLM 三层叠加，数据全程不出院。在 MIMIC-IV 出院小结上召回率比 Stanford/Penn 高 20 多个点，还同时提供高召回和高精确两种配置。做医疗 NLP 或医院信息化的团队可以直接拿来用。
DeidSharp 是一个开源 .NET 去标识化工具包，将可插拔检测器的结果组合起来：正则模式、临床 NER 模型和一个本地托管的 26B mixture-of-experts LLM，并在任何检测器报错时采取 fail-closed 策略。它在 PhysioNet 护理笔记（889 个 PHI span）和半合成的 MIMIC-IV-Note 出院小结（29,023 个 span）上，与 deid、Philter、Stanford/Penn de-identifier、OpenAI Privacy Filter 和 GLiNER 进行了对比评估，并使用配对的 patient-level bootstrap 置信区间。
在护理笔记上，召回优先的堆栈比 Stanford/Penn 找到 98.9% 的 PHI（+2.8, 95% CI 1.4–4.4），但精确率低 9.3 个百分点；不带 NER 的精确优先堆栈找到 96.5%，精确率达 83.2%，比 Stanford/Penn 高出 19.6 个百分点。在出院小结上，两个堆栈的召回率均比 Stanford/Penn 高出 20 个百分点以上（召回优先堆栈为 95.3%，+21.5），大部分提升来自生成的无类型名称；在有类型标签上，所有三种系统都超过了 99%。值得注意的是，LLM 是主要的检测器，NER 主要增加了地名；当要求 LLM 验证其自身的发现时，它会丢弃四分之一真正的 PHI。该工具包完全本地运行，因此 PHI 永远不会离开现场。
A local language model layered with rule-based and neural detectors for clinical text de-identification: an open-source toolkit and evaluation
Objectives: To measure whether a locally hosted large language model (LLM), stacked with deterministic and clinical named-entity recognition (NER) detectors, finds more protected health information (PHI) than open tools, and at what precision. Materials and Methods: DeidSharp, an open-source .NET toolkit, combines the findings of pluggable detectors (patterns, a clinical NER model, a local 26B mixture-of-experts LLM) and fails closed on any detector error. Configurations fixed on tuning halves were compared, under a prespecified protocol, with deid, Philter, the Stanford/Penn de-identifier, OpenAI Privacy Filter and GLiNER on reporting halves of PhysioNet's nursing notes (889 spans) and semi-synthetic MIMIC-IV-Note discharge summaries (29,023 spans), and on synthetic ASQ-PHI queries, with paired patient-level bootstrap intervals. Results: In the nursing notes, against Stanford/Penn, the recall-first stack found 98.9% of PHI (+2.8, 95% CI 1.4 to 4.4) at 9.3 points lower precision; the precision-first stack, without NER, found 96.5%, matching Stanford/Penn, at 19.6 points higher precision (83.2%). In the discharge summaries, both stacks led Stanford/Penn by more than 20 points (recall-first 95.3%, +21.5, 20.5 to 22.4), almost entirely on generated untyped names; on typed labels all three exceeded 99%. All four prespecified comparisons held. No configuration beat deid significantly on its home corpus; Philter found more discharge-summary spans than the recall-first stack at 23 points lower precision. Discussion: The LLM was the main detector and NER added place names; asked to verify its own findings, the LLM discarded a quarter of true PHI. Conclusion: A local LLM layered with deterministic detectors offers recall-first and precision-first operating points without sending text off-site.
本条情报归属 AITOP「论文」核心栏目，重点聚焦北美前沿实验室与硅谷科技巨头。在跨区域产业观察坐标中，该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」「中美大模型技术代差与商业化对照」研究主线，反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括：DeidSharp、LLM、PHI、临床文本，读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明：本文原始信息发布自一手权威信源「arXiv: OpenAI」（Tier-1 一手发布），具备高确定性参考价值。
