跳到正文

D003 敏感信息识别与去标识化

在本地识别并处理患者身份与其他敏感信息,保存去标识化结果和检查证据,为后续临床数据加工提供受控输入。

个人 Skill知识加工12 个文件浏览次数 0
开始使用

下载前阅读 SKILL.md 中的使用条件,并确认所用工具支持对应能力。

输入资料
Required: local source package path Recommended handoff artifacts: D001 source_manifest.json D002 schema_linkage_profile.json Example:D003 may run without D002, but identifier-role inference becomes less reliable and the… 阅读本节完整说明
交付内容
Always produce: phi_gate_report.json phi_gate_report.md deidentified_package/ for supported sanitized artifacts deidentified_manifest.json — source→sanitized artifact/status handoff privacy_review_queue.csv — review-requ… 阅读本节完整说明
  1. 获取完整包,保留配套文件的目录结构。
  2. 按照原文准备所需工具、依赖与输入资料。
  3. 阅读对应使用步骤与输入输出说明,在自己的环境中验证结果。

使用提示摘自原文,配套文件保留来源版本;实际运行条件以各包说明为准。

正文

获取原文件

README.md

v0.2.0 修复发布

补齐清单/画像引用,供最终门禁追溯。本地去标识规则通过不等于已取得患者资料的外发许可;自由文本仍需隐私审核。

D003 is the privacy boundary for the Haonan D-series medical-data processing pipeline.

It is intentionally conservative and uses only the Python standard library in v0.1. It does not claim HIPAA Safe Harbor or clinical-grade anonymization. Instead it creates a pseudonymized/redacted local package, audits residual high-confidence identifiers, and tells downstream skills whether content may become model-visible.

Quick start

python scripts/phi_gate.py \
  --input-path /path/to/source-package \
  --manifest /path/to/source_manifest.json \
  --profile /path/to/schema_linkage_profile.json \
  --output-root ./runs/haonan-d003

Then validate:

python scripts/validate_phi_report.py ./runs/haonan-d003/<package-id>/phi_gate_report.json

Self-test:

python scripts/self_test.py

Important limitation

Free-text de-identification is heuristic in v0.1. Sanitized free text therefore produces model_visibility_gate = review_required unless a future stronger local de-identification stage is added.

Whole-package network egress is always blocked.

Handoff artifacts

D003 also writes deidentified_manifest.json and privacy_review_queue.csv so downstream skills never have to guess which sanitized files are allowed, review-required, or blocked.

v0.1.1 clinical-field preservation

Clinical semantic fields such as diagnosis_name, test_name, drug_name, procedure_name and their Chinese equivalents are explicitly preserved. They must not be redacted merely because the field name contains name. Direct patient-name fields remain redacted.

当前版本:v0.2.1

新伪ID采用128 bit HMAC截断及a–p字母编码,保留角色域与运行级密钥,不生成可被手机号/身份证规则误匹配的长数字串。没有豁免任意数字或关闭隐私扫描。英文name标签添加单词边界。脱敏清单记录所有sanitized产物SHA-256;残留扫描读取错误阻断。

以SKILL.md顶部发布合同和references/repair-v0.2.1.md为准,历史自测描述不代表全面临床验收。

CONTACT HAONAN

联系haonan

医疗 AI · 产品与独立开发

刘浩南

关注医疗 AI 从 demo 到真实落地,聚焦产品化、商业化与院内工作流效率提升。

12年医疗产研经验,大湾区头部医疗AI公司前产品总监,独立开发者

医疗 AI 产品化医疗 AI 商业化院内工作流效率
刘浩南的微信联系二维码
微信扫一扫,联系 haonan