🛡️ Aegis — Medical Image De-identification

Upload a medical image (JPEG/PNG) containing burnt-in Protected Health Information (PHI) such as patient names, dates, hospital names, or IDs.

Aegis detects and redacts PHI using a 3-layer classification pipeline:

  1. Clinical Allowlist — device manufacturers, imaging modes (preserved)
  2. PHI Heuristics — date-ID combos, hospital names (redacted)
  3. Stanford NER Model — semantic PHI classification with F1 ≥ 97.9% (redacted)

Clinical measurements, probe IDs, and technical annotations are preserved.

🔗 GitHub Repository · 📄 DICOM PS3.15 & HIPAA Safe Harbor compliant

📤 INPUT

🔒 OUTPUT

Upload an image and click the button above to see the de-identification report.

Built with Aegis · MONAI · Stanford NER · EasyOCR

© 2026 Lakshmi Mahabaleshwara · Apache 2.0