Facial feature removal in magnetic resonance imaging scans of adults with Down syndrome: A de‐facing methodological study
Jason K. Russell, Michael S. Rafii, Walter K. Kremers, Scott A. Przybelski, Prashanthi Vemuri, Ronald C. Petersen, Jonathan Graff‐Radford, Clifford R. Jack, Christopher G. SchwarzAbstract
INTRODUCTION
Face identification algorithms have increased the risk of study participants being identified through different neuroimaging modalities. De‐facing algorithms can de‐identify neuroimages; however, none are validated in individuals with Down syndrome (DS) – a population at high risk of Alzheimer's disease and the focus of numerous Alzheimer's disease cohort studies.
METHODS
Overall, 37 adults with DS were parcellated using FreeSurfer before and after de‐facing with mri_reface . A group of neurotypically developed individuals balanced on age, sex, and magnetic resonance imaging scanner manufacturer served as controls.
RESULTS
De‐facing produced no effect on cortical thickness or volumetrics in adults with DS compared to controls or regional discrimination based on an area under the receiver operating characteristic curve analysis after correcting for multiple comparisons.
DISCUSSION
FreeSurfer‐derived volumetric and cortical thickness measures changed minimally following de‐facing with mri_reface in adults with DS and are unlikely to influence study outcomes. De‐facing should be considered prior to data sharing in studies of adults with DS.