电子医疗文档的流式动态脱敏实现研究.pdfVIP

智慧医院建设与实践

ConstructionandPracticeofSmartHospital

电子医疗文档的流式动态脱敏实现研究

何剑虎伊胜月宋丽莹

【摘要】目的：实现电子医疗文档在共享时的脱敏处理，保护患者隐私。方法：构建一个集成多种机器学习模

型的医疗数据词法分析器，整理医疗健康领域的中文分词、词性标注和命名实体识别语料库，利用隐马尔可夫、

条件随机场等自然语言处理技术和内置敏感信息特征库识别电子医疗文档中的敏感信息，并通过结果集流式处

理技术实现动态脱敏。结果：算法模型在处理常规个人敏感信息时效果较好，个人敏感信息的发现与脱敏平均

耗时在毫秒级别。结论：自然语言处理结合敏感信息特征库的方法可实现非结构化电子医疗文档敏感信息的识

别与实时脱敏。

【关键词】电子医疗文档；敏感信息识别；动态脱敏；自然语言处理；流式处理

Doi:10.3969/j.issn.1673-7571.2025.04.010

【中图分类号】R197.3；R319

Researchontherealizationofdynamicdesensitizationofelectronicmedicaldocumentsdata

throughstreamingprocessing

HEJianhu,YIShengyue,SONGLiying.WomensHospitalSchoolofMedicineZhejiangUniversity,Hangzhou310006,

ZhejiangProvince,China

Correspondingauthor:HEJianhu,Email:hejh@

Toachievedesensitizationofelectronicmedicaldocumentsduringsharing,andtoprotect

【Abstract】Objective

patientprivacy.MethodsAlexicalanalyzerformedicaldataintegratingmultiplemachinelearningmodelswas

constructedtosortoutChinesewordsegmentation,part-of-speechtagging,andnamedentityrecognitioncorporain

thefieldofmedicalandhealthcare.Sensitiveinformationinelectronicmedicaldocumentswasidentifiedbyusing

naturallanguageprocessingtechnologiessuchasHiddenMarkovModelsandConditionalRandomFieldsandbuilt-in

sensitiveinformationsignaturelibrary,anddynamicdesensitizationwasrealizedthroughresultsetstreamingprocessing

technology.ResultsThealgorithmmodelhasagoodeffectontheprocessingofroutinesensitivepersonalinformation,

withanaveragetimeofdetectionanddesensitizationofsensitivepersonalinformationwasmilliseconds.Conclusion

Themethodofnaturallanguageprocessingwithsensitiveinformationsignaturelibrarycanrealizethereco