Face de-identification (head defacing)

Auto-detect and redact facial features in CT/MR head scans.

Problem

Head CT and MR scans contain 3D facial surface data that can be reconstructed to identify a patient through facial recognition. Metadata anonymization doesn’t address this — the facial features are embedded in the pixel data itself. You need to automatically detect and deface the facial regions in head scans before the data leaves your institution.

Steps

1. Open your head CT/MR DICOM file (⌘O) and switch to Anon mode (⌘2).

2. Activate auto-defacing — click Auto-Deface in the Anonymization panel. The tool analyzes the current frame’s anatomy and proposes defacing regions.

3. Review the proposals — three regions are registered automatically:

   • Face (the primary facial surface)
   • Left Ear
   • Right Ear

   The redaction regions counter should show (3).

4. Apply & Export — click the Apply & Export button to write the defaced DICOM file.

Expected Result

• After clicking Auto-Deface, exactly 3 redaction regions are proposed (Face, Left Ear, Right Ear).
• The regions are overlaid on the viewport as colored bounding boxes.
• Apply & Export produces a valid DICOM file with the facial regions blacked out in the pixel data.
• The defacing targets the soft tissue and bone surfaces of the face, ears, and anterior skull — the brain parenchyma and intracranial structures remain untouched.
• This feature is designed for head CT and MR scans. It may not produce meaningful proposals for other anatomical regions or modalities.