Identification of Anonymous MRI Research Participants with Face-Recognition Software

Abstract

A, Volume rendering from the original nonedited CT scan. B, Volume rendering from the edited CT image does not display any unique distinguishing facial features. smoothed using a Gaussian smoothing method with a standard deviation of 5 mm. Then, a surface model was generated from the dilated and smoothed segmentation to replace the original air and skin voxels in the original CT scan. This method created a new DICOM volume with the intracranial anatomy preserved (Figure; Movie [online]).