Face De-Identification Service for Neuroimaging Volumes

Abstract

Digital medical imaging is a fundamental tool for improving medical practice workflows and supporting clinical diagnosis. Nowadays, healthcare institutions are usually very supported by information and communication systems that meet regular practice requirements. However, the usage of those platforms in collaborative, research and educational scenarios faces several problems. One of the key issues is related with patient data privacy, namely with concerns related with the visual anonymization of studies. In the neuroimaging field, this subject is more complex since, even after removing the patient's information from the images meta-data or burned in the pixel data, it is still possible to identify the patients through 3D reconstruction of the volume. This article proposes and describes the implementation of an end-user service that allows neuroimages facial de-identification of CT volumes, being fully interoperable with production repositories. The solution was validated using a public dataset and made available to the community through its integration with an open source archive server.