TY  - JOUR
AU  - Ye, Kai
PY  - 2025
DA  - 2025/11/12
TI  - Hybrid Large Language Model and Rule-Based Framework for Automated PHI De-Identification in Clinical Notes
JO  - ICCK Transactions on Emerging Topics in Artificial Intelligence
T2  - ICCK Transactions on Emerging Topics in Artificial Intelligence
JF  - ICCK Transactions on Emerging Topics in Artificial Intelligence
VL  - 3
IS  - 1
SP  - 1
EP  - 8
DO  - 10.62762/TETAI.2025.518010
UR  - https://www.icck.org/article/abs/TETAI.2025.518010
KW  - PHI de-identification
KW  - clinical NLP
KW  - large language models
KW  - hybrid systems
KW  - parameter-efficient fine-tuning (PEFT)
KW  - electronic health records
KW  - privacy preservation
KW  - retrieval-augmented generation (RAG)
KW  - rule-based NLP
KW  - biomedical text processing
AB  - The growing demand for secondary use of electronic health records (EHRs) in clinical research has amplified the importance of effective de-identification of protected health information (PHI) to comply with privacy regulations such as HIPAA. Manual annotation remains error-prone, time-consuming, and inconsistent across healthcare institutions, while existing automated systems often face trade-offs between accuracy, interpretability, and computational cost. This study proposes a novel hybrid de-identification framework that integrates neural, statistical, and rule-based approaches to achieve high recall, operational efficiency, and deployment feasibility in real-world healthcare settings.
SN  - 3068-6652
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Ye2025Hybrid,
  author = {Kai Ye},
  title = {Hybrid Large Language Model and Rule-Based Framework for Automated PHI De-Identification in Clinical Notes},
  journal = {ICCK Transactions on Emerging Topics in Artificial Intelligence},
  year = {2025},
  volume = {3},
  number = {1},
  pages = {1-8},
  doi = {10.62762/TETAI.2025.518010},
  url = {https://www.icck.org/article/abs/TETAI.2025.518010},
  abstract = {The growing demand for secondary use of electronic health records (EHRs) in clinical research has amplified the importance of effective de-identification of protected health information (PHI) to comply with privacy regulations such as HIPAA. Manual annotation remains error-prone, time-consuming, and inconsistent across healthcare institutions, while existing automated systems often face trade-offs between accuracy, interpretability, and computational cost. This study proposes a novel hybrid de-identification framework that integrates neural, statistical, and rule-based approaches to achieve high recall, operational efficiency, and deployment feasibility in real-world healthcare settings.},
  keywords = {PHI de-identification, clinical NLP, large language models, hybrid systems, parameter-efficient fine-tuning (PEFT), electronic health records, privacy preservation, retrieval-augmented generation (RAG), rule-based NLP, biomedical text processing},
  issn = {3068-6652},
  publisher = {Institute of Central Computation and Knowledge}
}