Implementation and validation of face de-identification (de-facing) in ADNI4.

Area of Science:

• Neuroimaging
• Data Privacy
• Medical Informatics

Background:

• Technological advancements increase re-identification risk for de-identified brain images.
• The Alzheimer's Disease Neuroimaging Initiative (ADNI) is a key public source of brain imaging data.
• Protecting participant privacy in neuroimaging datasets is critical.

Purpose of the Study:

• To evaluate and select a face-deidentification method for ADNI brain images.
• To ensure participant privacy is maintained without compromising data integrity.
• To implement a robust de-facing workflow for ADNI data.

Main Methods:

• An expert committee evaluated 11 de-identification algorithms.
• Four algorithms underwent formal testing for efficacy and impact on brain measurements.
• The mri_reface algorithm was selected based on reliability and practical considerations.

Main Results:

• De-facing effects on brain measurements were minimal and comparable across tested methods.
• mri_reface demonstrated superior reliability and practical advantages.
• The chosen method was validated for robustness and effectiveness on ADNI sequences.

Conclusions:

• ADNI implemented "de-facing" of MRI and PET images starting with ADNI4.
• mri_reface is used to protect privacy by altering facial imagery.
• Rigorous validation confirmed negligible effects on brain measurements and complete face removal.