Researchers at Rochester, Minn.-based Mayo Clinic used commercial facial recognition software to identify pictures of volunteer research participants to unidentified MRI scans, with the findings drawing attention to patient privacy concerns, according to The Wall Street Journal.
The study was published in the New England Journal of Medicine. Mayo Clinic recruited 84 volunteers ages 34 to 86 who had received an MRI scan of the head during a clinical study. Study participants were photographed from five angles, and researchers reconstructed an image of their face using the MRI scans.
Mayo Clinic researchers sought out this study after noticing how clear images used to look at brain structures were. Christopher Schwarz, a Mayo Clinic computer scientist and lead author of the study, thought that he could recognize a patient based on just the MRI scans because of their clarity.
Facial recognition software from Microsoft Azure was able to correctly match 70 of the 84 images. For 10 other images, the software ranked the correct photo among its top five candidates, reports WSJ.
Mayo Clinic researchers want to develop a solution for the privacy issues. However, blurring or removing data from MRIs could limit the quality of the scans.