Re-Identification of Anonymized MRI Head Images with Publicly Available Software: Investigation of the Current Risk to Patient Privacy

A study examined the privacy risks of using facial recognition software (FRS) on anonymized MRI head images. Involving 4 participants, the research showed that 50% could be identified using 3D renderings from MRI scans, matched against a database of over 800 million images. Factors like head orientation and hairstyle influenced accuracy, but the absence of hair in MRI scans posed challenges. The study found that using the mri_deface algorithm prevented recognition, suggesting that standardized de-facing protocols could mitigate privacy risks. The findings highlight the need for improved data protection measures in clinical trials to balance data sharing with privacy protection.