Clinical DeIdentification

The Clinical De-Identification model is designed to recognize and anonymize PHI in English-language clinical notes. It employs state-of-the-art natural language processing techniques to detect sensitive information such as patient names, addresses, medical record numbers, and other identifiers. Once identified, the PHI is effectively masked/obfuscated, rendering the text safe for broader use while maintaining its informational integrity.

<p><br/></p>

Key Features:

- The model is finely tuned to identify a wide range of PHI elements in medical texts, ensuring comprehensive de-identification.

- The de-identification process aligns with HIPAA and other healthcare privacy regulations, aiding in legal compliance and data protection.

- Ideal for research, analytics, and training purposes, this model enables the safe utilization of medical texts without compromising patient privacy.

<p><br/></p>

This model is a useful asset in the healthcare and research sectors, where the protection of patient privacy is paramount. It allows for the ethical and legal use of valuable medical data, promoting research and analysis while upholding the highest standards of data privacy and security.

<p><br/></p>

Covered entities: AGE, CONTACT, DATE, ID, LOCATION, NAME, PROFESSION, CITY, COUNTRY, DOCTOR, HOSPITAL, IDNUM, MEDICALRECORD, ORGANIZATION, PATIENT, PHONE, PROFESSION, STREET, USERNAME, ZIP, ACCOUNT, LICENSE, VIN, SSN, DLN, PLATE, IPADDR, EMAIL entities.

<p><br/></p>

Available `masking_policies` are : `masked` (default one), `obfuscated`, `masked_fixed_length_chars` and `masked_with_chars`.

搜集汇总

背景与挑战

背景概述

Clinical DeIdentification模型采用先进NLP技术识别英语临床笔记中的敏感信息（如姓名、地址、病历号等），并按照HIPAA等法规进行匿名化处理。该模型支持多种掩码策略，覆盖28类PHI实体，适用于医疗研究和数据分析场景，确保数据可用性的同时保护患者隐私。