NIH | National Cancer Institute | NCI Wiki  

The Center for Biomedical Informatics and Information Technology of the National Cancer Institute (NCI) presented a virtual Medical Imaging De-Identification (MIDI) workshop focused on public sharing of imaging data.

The primary emphasis of the workshop was on medical images with accompanying data elements, especially those in formats in which the data elements are embedded, particularly DICOM.

The goals of the two-day workshop were to:

  • Share best practices and recommendations for medical imaging de-identification, as identified by the MIDI Task Group convened by the NCI.
  • Learn about approaches to conventional image de-identification in the United States, the European Union, and Canada.
  • Discuss approaches to image de-identification by industry.
  • Explore the roles of statistical risk analysis, de-facing, and AI in de-identification.

Recordings

DateSlidesRecording
5/22/2023

Day 1 Recording

5/23/2023Day 2 Recording

Slide Presentations

The workshop agenda and speaker bios are on the workshop website

Some documents on this page are not Section 508 compliant. To receive a compliant document, please email NCI CBIIT MIDI Group.

SpeakerPresentation Title

Slides

Day 1


Keyvan Farahani, PhD, National Heart, Lung, and Blood Institute, National Institutes of HealthWelcome & Introduction
Session 1: Report of the MIDI Task Group - Best Practices and Recommendations


David Clunie, MBBS, PixelMedReport of the MIDI Task Group

Session 2: Tools for Conventional Approaches to De-Identification

Chairperson, Fred Prior, PhD



Fred Prior, PhDSetting the Stage
Michael Rutherford, MS, University of Arkansas for Medical SciencesThe Tools of TCIA: Standardizing Zero-Tolerance De-identification
Stephen Moore, MS, Washington University School of Medicine in St. LouisXNAT Platform: Image De-identification

Session 3: International Approaches to De-Identification

Chairperson: William Parker, University of British Columbia



William Parker, MDMedical Data De-ID -
A Canadian Perspective
Parker Slides
Haridimos Kondylakis, PhD, Institute of Computer Science, Foundation of Research & Technology (FORTH)Data Infrastructures for AI in Medical Imaging: A report on the experiences of five EU projects
Christian Ludwigs, MSc, Aigora GmbHLegal framework and best practices for medical image de-identification in the EU

Session 4: Industry Panel on Image De-Identification

Chairperson: Juergen Klenk, PhD, Deloitte Consulting



Juergen Klenk, PhDIntroductory Remarks to the Industry Panel
Bob Lou, MD, GoogleMedical imaging de-identification on both images and text using AI models
Lawrence (Tony) O’Sullivan, MS, IBISOptimizing and Automating Radiology
Data De-identification Workflows
Dan Marcus, PhD, FlywheelThe Flywheel Platform for Intelligent Image Anonymization
Jiri Dobes, PhD, John Snow LabsAutomated Medical Data De-Identification and Obfuscation
Abraham Gutman, MS, AG Mednet, Inc.Advances in Medical Imaging De-Identification and the Impact of Regulatory Constraints
Day 2

Session 5: Pathology Whole Slide Image De-Identification

Adam Taylor, PhD, Sage BionetworksPathology Whole Slide Image De-Identification
Tom Bisson, PhD, Charité Universitätsmedizin BerlinAnonymization of Whole Slide Images in in Histopathology for Research and Education
David Gutman, MD, PhD, Emory UniversityImage DePHI and the DSA: Open Source tools for Histology Image DeIdentification
Session 6: De-FacingDe-Facing
Ying Xiao,PhD, Hospital of the University of PennsylvaniaMedical Image De-Facing and Clinical Research Data Sharing
Christopher Schwarz, PhD, Mayo ClinicFace Recognition and De-Identification of Research Brain Images with mri_reface
Douglas Greve, PhD, MGH/HarvardMIDEFACE: Minimally Invasive Defacing
Session 7: The Role of AI in Image De-Identification

Judy Gichoya, MD, Emory University
George Shih, MD, Weill Cornell Medical CollegePixel De-Identification Using AI
Adrienne Kline, MD, PhD, Northwestern UniversityPyLogik: An open-source resource for medical image de-identification
Session 8: NCI MIDI Datasets and Pipeline

Keyvan Farahani, PhDThe Medical Image De-Identification Initiative (MIDI)
Fred Prior, PhDSynthetic Data for De-Identification Testing -
The MIDI Datasets
Ben Kopchick, PhD, Deloitte ConsultingBuilding a cloud-based MIDI pipeline