This is the home page for the Medical Imaging De-Identification (MIDI) project.

MIDI Project Goal

Drive consensus, provide guidance, and promote best practices for medical image de-identification for various imaging informatics applications in academic and industrial research.

MIDI Project Objectives

The MIDI projects described below focus on medical imaging in radiology and the DICOM standard, as the most basic, and yet the most prevalent need or use case.  Future extensions of these projects could address imaging de-ID in other domains such as digital pathology (whole slide imaging) and multi-spectral imaging.  There will be many NCI use cases for these projects, including IDC, potentially TCIA, SBIR imaging de-ID contract product validation, benchmarking of de-ID community tools through challenges, and promotion of data sharing in the research community.

Project 1: Medical image de-ID assessment (MIDIA) reference datasets

• Goal: To design and create a comprehensive imaging dataset using real patient image data, anonymized, and injected with artificial PHI and identifiable information in conventional and non-conventional DICOM fields in the image file header and in the pixel data.
• Collaborators: Fred Prior, PhD (TCIA/UAMS) and David Clunie, MD (PixelMed)
• Project period: June-Aug 2020
• Funding source: Possibly through existing CBIIT contracts in Imaging Informatics for Precision Medicine, including a DICOM contract with D. Clunie.
• Current status: Based on preliminary, and separate, discussions with Fred Prior, David Clunie, they have been interested in developing image de-ID validation datasets. However, to date, no concerted effort has been made. 

Project 2: MIDI Tools and Pipeline

• Goals: To integrate best of the class image de-ID tools developed by Google Health for scalable applications, and to develop a pipeline for assessment (validation) and application of de-identification tools to imaging datasets.
• Collaborators: Deloitte Consulting
• Project period: June 2020 – October 2020
• Funding source: CBIIT funds through the GCP STRIDES program
• Current status: CIT and GCP STRIDES have confirmed the availability of GCP STRIDES mechanism toward Professional Services Development.  CBIIT has approved the availability of STRIDES credits to complete this project.  There is a possibility of co-funding by NHLBI as they have expressed interest in this project.

Project 3: MIDI Working Group

• Goal: To develop conventions, guidance, and Best Practices for image de-ID for imaging informatics applications in cancer diagnosis and therapy (i.e., radiotherapy)
• Potential members: To be determined, but will likely include members from NCI (CBIIT, CIP), TCIA, IDC, Deloitte Consulting, industry (e.g., Google Health, Amazon Web Services, PixelMed, etc.), non-profits, including the American College of Radiology (ACR), the Radiological Society of North America (RSNA), the Society for Imaging Informatics in Medicine (SiiM), the American Association of Physicists in Medicine (AAPM), and the American Society for Therapeutic Radiation Oncology (ASTRO), NIBIB, NHLBI, and FDA
• Project period: July 2020 – June 2022
• Funding source: volunteer effort. Periodic Working Group meetings will be done online and possibly in-person once a year at a member society’s annual meeting.
• Current status: NCI, IDC, TCIA, PixelMed, and ACR have expressed interest in formation of a working group (or forum) to address image de-ID methods. It may be an effective approach to initiate the working group upon initiation of Project 1, and possibly the NCI-ACR MOU for collaboration.