Edit this page

NA-MIC Project Weeks

Back to Projects List

NCI Imaging Data Commons - user support and platform development

Key Investigators

Presenter location: In-person

Project Description

NCI Imaging Data Commons is a cloud-based environment containing publicly available cancer imaging data co-located with analysis and exploration tools and resources.

IDC provides a growing amount of publicly available cancer imaging data (>65TB at the moment, radiology and digital pathology, including images, annotations, analysis results and clinical data) curated in the cloud to support highly efficient access and to simplify analysis.

Objective

  1. Raise awareness about IDC, help users, collect feedback to help prioritize future development.
  2. Identify robust AI models that can be applied to IDC data to enrich IDC with annotations.
  3. Work on various issues related to the development of IDC platform and related software tools.

Approach and Plan

  1. Interact with current and prospective users to answer questions and collect feedback.
  2. Support any project that has a need for public datasets available for testing, cloud-based notebook implementations of the analysis, scaling up analysis to large cohorts within IDC.
  3. Work on priority aspects of the project: maintenance and improvement of SlicerIDCBrowser and idc-index, improvements of the documentation and other learning materials
  4. Improve/simplify access to the NLST/TotalSegmentator analysis results.
  5. Work on maintenance of dcmqi priority issues: https://github.com/QIICR/dcmqi/issues/489, python wrapper API
  6. MRTotalsegmenator SCT codes - Andras
  7. DCMTK upgrade in Slicer - JC

Progress and Next Steps

  1. Update MHub+IDC tutorial in how it accesses IDC.
  2. Prepared initial version of the query to extract processing steps for slide microscopy (SM) images using DICOM metadata (https://github.com/ImagingDataCommons/idc-index-data/pull/30). When completed, this will allow selecting SM images by embedding method, staining (H&E), and fixative without using BigQuery, and with queries of significantly lower complexity as compared to querying full index.
  3. Implemented new feature in the dcmqi converter that allows including into DICOM SEG references to the segmented images when geometry of the segmentation is different from the image (e.g., when segmentation was done on the slices orthogonal to the segmented image) (https://github.com/QIICR/dcmqi/issues/489). Lacking this feature, ReMIND collection encoded images that are disconnected from the segmented MR images.
  4. Mapped model-specific segmentation labels for OMAS and TotalSegmentator to SNOMED-CT (related PRs https://github.com/wasserth/TotalSegmentator/pull/324 and https://github.com/wasserth/TotalSegmentator/pull/325).
  5. Presented IDC updates at the Thu breakout session (see notes and references in this document).
  6. Reviewed beta (aka pita) release of the pydcmqi python wrapper of dcmqi prepared by Leo. pydcmqi aims to simplify pythonic access to dcmqi functionality.

Illustrations

Summary of IDC content as of data release v18

Background and References