(This meeting is not online.)

DICOM Sub-Group Call Notes

July 13, 2026

Attendees:

• Kincaid, Heather L (US 342C-Affiliate)
• Fu, Sidney (NIH/NCI) [E]
• Ozyoruk, Kutsev Bengisu
• Kelly, Sean C (US 342C-Affiliate)
• Crichton, Daniel J (US 3410)
• Page-Lester, Stephanie
• Madabhushi, Anant
• Kohaar, Indu (NIH/NCI) [E]
• Salami, Simpa (Simpa)
• Lopez, Camden
• Hsu, William
• Yemparala, Gireshbabu
• Reid, Suzanna L
• Erin Fowler (Unverified)
• Balagurunathan, Yoga
• 18137454673 (Unverified)
• Savannah C. Partridge
• Stoyanova, Radka
• Dahlgren, Jackie
• Mahabal, Ashish
• Schabath, Matthew B

Discussion Topics

• DICOM Validation Progress and Requirements Finalization: progress made on DICOM validation, including updates to core tags, extensions for PET and RT, and the implementation of validation scripts, with the goal of finalizing requirements for both legacy and new data in LabCAS.
• Core Tag Updates: The core DICOM tags were updated based on previous calls, and extensions for PET and RT were added to support LTP 2 and pdoc images, ensuring compatibility with various modalities.
• Versioning of Requirements: The team established two versions of requirements: version one for legacy images with less strict criteria, and version two for new images with stricter requirements, allowing for phased implementation and minimizing site burden.
• Validation Script Implementation: A validation script was developed to check for required tags, proper formatting, and PHI in DICOM headers, with reconciliation reports generated for validation and PHI findings, and pilot testing conducted at UCLA.
• UID Regeneration: Heather described collaboration with Dr. Hsu and Dr. Lenburg to regenerate missing UIDs for a site, eliminating the need for re-uploading images from the radiology department.
• Training and Portal Updates: Training materials for PMRI study coordinators were provided and posted on the portal, and updated header tags and de-identification processes were made available for site review.
• PHI Detection Improvements: William reported that UCLA’s pilot PHI report produced numerous false positives, particularly from social security number patterns. Sean and Daniel explained that the detection algorithm uses pattern recognition and is being refined to reduce false positives; updated reports will be distributed, and local deployment of the validation tool is planned to shorten review cycles.
• Tool Deployment Plans: Daniel shared the intention to push the validation tool to sites for local use, reducing data upload requirements and cycle times, with Sean actively working on this deployment.
• Grouping Errors by Series: Heather mentioned efforts to organize validation errors by series rather than individual files, making review easier for sites.
• Blinding Requirements for DICOM Reference Sets: Camden and the DMCC team outlined evolving blinding requirements for DICOM reference sets, proposing unique patient IDs per imaging study and fixed arbitrary dates to prevent longitudinal linkage, with input from Suzanna, Radka, William, Savannah, and others.
• Validation Summary Reports for pMRI and LTP 2: Heather presented validation summary reports for pMRI and LTP 2, detailing missing required tags, critical errors, warnings, and PHI findings, with input from Radka and Savannah on the necessity of specific tags such as spacing between slices and diffusion B values.
• Common Missing Tags: Heather identified acquisition time, content time, series description, and spacing between slices as the most commonly missing tags in pMRI, with series description retained as required based on prior decisions.
• Tag Requirement Discussion: Radka and Savannah discussed the importance of spacing between slices and diffusion B values, noting modality-specific relevance and vendor variability, and recommended retaining these tags where possible.
• Validation Report Structure: Validation reports include percentages of missing tags by site, critical errors, warnings, and PHI findings, with warnings not sent to sites and critical errors prompting site review and potential re-upload.
• CTDL Vol Tag Issue: Heather noted that CTDL Vol, part of the CT extension, was frequently missing in LTP 2, and sought input on whether it should be required or optional, with plans to update validation scripts for nested tag parsing.
• Validation Approach for New pMRI Uploads: The new pMRI uploads will be validated using version one requirements, with the team planning to reassess and potentially adopt stricter criteria after field testing and user feedback.
• Discussion of Intermediate Repository for Longitudinal Data: Heather, Camden, and Suzanna considered the creation of an intermediate repository preserving longitudinal information, agreeing to further discuss its purpose and potential future needs for public data release.

Pending Decisions

• Unique Patient ID Policy: Camden proposed assigning a unique patient ID to each imaging study, preventing linkage between multiple studies from the same patient and thereby blinding users to longitudinal information.
• Date Shifting Strategy: The DMCC suggested setting all study dates to a fixed arbitrary date, such as January 1, 1960, to further blind users to temporal information and ensure clarity that the date is not real.
• Handling Longitudinal Data: Discussion addressed the rare cases where longitudinal linkage is needed for analysis, with the default policy being to blind images unless justified by specific study requirements, and metadata provided separately if needed.
• Metadata Provision: Savannah and William confirmed that metadata, including exam-level IDs and dates, can be provided separately by the DMCC for analysis, and most investigators do not require this information in the image header.
• DICOM Reference Set Blinding: Unique patient IDs per study and a fixed date of January 1, 1960, will be used for all images in the reference set.
• Centralized De-identification: JPL will perform centralized stripping of institution identification for the reference set to ensure consistency across sites.

Action Items

• Review of Portal Materials: Send links to the posted header tags and de-identification process materials on the portal for review by participants. (Heather)
• pMRI Image Examples: Request permission from DMCC to send some images with missing diffusion B values and Spacing Between Slices to Radka for review to determine if validation code needs to be modified. (Heather)
• Validation Report Distribution: Send site-level validation summary reports (not including warnings) to relevant sites for review and feedback. (Heather)
• Blinding Requirements: Provide JPL with the final blinding criteria for the DICOM reference sets. (DMCC)

When

Los Angeles

July 13, 2026, 10 a.m.

Denver

July 13, 2026, 11 a.m.

Chicago

July 13, 2026, noon

New York

July 13, 2026, 1 p.m.

UTC

July 13, 2026, 5 p.m.