A fundamental requirement for sharing medical images (DICOM format), particularly through public image repositories, is the deID of protected health information (PHI) and personally identifiable information (PII) from the DICOM header and the pixel matrix of all images. In the context of the MIDI-B challenge, deID refers to adherence to the US HIPAA Privacy Rule safe harbor method, DICOM Standard PS3.15 (Attribute Confidentiality Profile), and Best Practices described in the TCIA Submission Overview Page (see sections: Radiology Curation Steps and DICOM De-identification Details). In this challenge, we will use images from the MIDI dataset, a small portion of which is described in Rutherford, _et. al._. The MIDI dataset is a large and diverse image set from multiple modalities and submitting sites, as hosted in TCIA and IDC, infused with synthetic identifiers (PHI/PII). Automated methods for imaging de-identification that preserve the research utility of the data, considered as best practice, are desirable.

 

Participate in this challenge via the official challenge website: https://www.synapse.org/Synapse:syn53065760/wiki/627878

  • No labels