AI for Protected Health Information Detection
Securing Medical Images through Automated PHI Detection
This research evaluates an AI-based system to detect and protect sensitive patient information in medical images, addressing critical healthcare data protection requirements.
Key Findings:
- A three-component pipeline (text localization, extraction, and analysis) effectively detects PHI in medical images
- Tested across four modalities: CT, X-ray, bone scan, and MRI
- Evaluated multiple vision and language models for optimal PHI detection performance
- Created a specialized dataset with simulated PHI across eight categories
Why This Matters: Healthcare organizations can leverage these AI systems to automate compliance with privacy regulations, reduce manual review processes, and minimize the risk of exposing sensitive patient information in shared medical images.