Sharing standardized image-derived data in computational pathology using DICOM
Daniela P. Schacherer, Christopher P. Bridge, David Clunie, Igor Octaviano, André Homeyer, Markus D. Herrmann, Olivier Gevaert, Tabita Ghete, et al.
cs.CV · cs.AI · cs.LG
Quick verdict
Claims DICOM standardizes sharing of image-derived pathology data; five IDC collections and tooling support it, but benefits are asserted, not measured.
Summary
This data-infrastructure paper describes the first systematic effort to publish image-derived computational pathology data—annotations, segmentation masks, quantitative maps—in standard DICOM within the NCI Imaging Data Commons (IDC). The team converted five heterogeneous datasets: automated nuclei contours for 6065 TCGA slides across 14 cancer types (~7 TB as MBSA plus binary SEG), expert and model tissue typing for 96 rhabdomyosarcoma slides (SR TID1500 plus fractional SEG), patch-level TIL maps for 8367 slides across 23 TCGA collections (SEG), GBM aggressiveness maps (PM), and expert bone-marrow ROI/cell annotations (MBSA). All objects passed DICOM conformance validation (dciodvfy, dcentvfy, PixelMed SR validator) before ingestion and Slim-viewer visualization. Beyond the releases, the work substantially extended highdicom (tiling, pyramids, labelmap SEGs, MBSA parsing; versions 0.21.0–0.28.0) and Slim (non-matching resolution support, ROI clustering). The paper distills practical guidance: prefer tiled-full over tiled-sparse, image-relative coordinates, MBSA over SR for bulk ROIs; labelmap SEGs measured 14–48× smaller than uncompressed binary SEGs. Main limitations: benefits such as searchability and interoperability are demonstrated qualitatively (dashboard, notebooks) rather than measured; validation covers standard compliance, not conversion fidelity; comparative claims against OME formats are unevaluated; several conversions required undocumented semantics recovered via personal communication.
Review complete
Assessment at a glance
5
risk flags identified
● high 0
● medium 3
● low 2
4 strengths · 5 weaknesses · 5 author questions
AI text signal = Binoculars zero-shot detector score over 20 sampled passages (Fast-DetectGPT co-reported) — absence does not prove human authorship, and a flag does not imply misconduct.
Levels reflect verification complexity — not a quality score. Evolve never rates your paper.
Key claims under review
- DICOM's SEG, PM, SR (TID1500), and MBSA IODs can encode the breadth of pathology image-derived data, demonstrated by five public IDC collections spanning raster/vector representations, manual and AI provenance, and coarse-region to subcellular scales.
- This constitutes the first public release of large-scale scientific image-derived pathology data in DICOM format.
- Harmonization into DICOM yields unambiguous identification and formal linkage, coded semantics enabling cross-collection structured querying, integrated context/provenance, and uniform programmatic access.
- Practical defaults for future converters are tiled-full representation, image-relative coordinates, MBSA for bulk ROIs, and labelmap SEGs (measured 14–48× smaller than uncompressed binary SEGs in their cases).
Strengths
Fills a genuine gap at real scale
First public release of large-scale image-derived pathology data in DICOM: five IDC collections (Table 2) spanning 6065 nuclei-annotated slides/6970.57 GB (Pan-Cancer-Nuclei-Seg-DICOM, v19), 96 RMS slides (SR TID1500 + fractional SEG), 8367 TIL slides across 23 collections, GBM360 PMs, and bone-marrow MBSAs in v24 (2026), covering raster/vector and manual/AI provenance.
Durable open-source contributions
highdicom 0.21.0→0.28.0 gains (SEG efficiency, tiled-full/sparse tiling, pyramid creation, labelmap SEG type, MBSA parsing, CIELab utilities) and Slim fixes (SEG/PM at resolutions absent from the source pyramid; spatial clustering of hundreds of thousands of ROIs) remove concrete adoption blockers for anyone repeating this work.
Evidence-grounded engineering guidance
Strong FAIR mechanics end-to-end
Weaknesses
Benefits asserted, not measured
Validation is conformance-only
Stakeholder perspective on the format comparison
Workaround encodings reduce out-of-box interoperability
Table 2 arithmetic left to the reader
Full Review Report
$1.99 · one-timeUnlock the complete preliminary analysis Evolve produced for this paper.
Report preview
Scope & Claims
This is a data-infrastructure and standards-adoption paper, not an algorithmic contribution. A consortium spanning the IDC team, Fraunhofer MEVIS, Stanford, Erlangen, and the core DICOM tooling ecosystem (PixelMed, highdicom) makes three central claims: (1) DICOM's SEG, PM, SR (TID1500), and MBSA IODs can encode the breadth of pathology image-derived content; (2) this is the first public release of large-scale image-derived pathology data in DICOM, delivered as five IDC collections (Table 2: Pan-Cancer-Nuclei-Seg-DICOM, 6065 slides, 6970.57 GB, first in v19/2024; through BoneMarrowWSI-PediatricLeukemia in v24/2026); (3) harmonization yields unambiguous identification and linkage, coded semantics enabling cross-collection structured queries, provenance integration, and uniform programmatic access. The claims are largely supported by construction — the data exist, pass validators, and render in Slim — but comparative benefits are argued, not measured.
The remaining sections cover Methodology Assessment, Statistical Evidence, Results & Reproducibility, Limitations & Risks, and Verdict & Recommendations.
- All 4 strengths & 5 weaknesses, analyzed in detail
- 5 risk flags with severity, section & reviewer questions
- Methods & statistical evidence assessment
- 5 questions a peer reviewer would ask
- Full six-dimension evaluation report & recommendations
One-time purchase · this paper only
One-time purchase, this paper only. AI-generated preliminary analysis — not peer review.
Frequently asked questions
What does this paper actually release?
Five collections of image-derived pathology data (nuclei contours, tissue-type annotations, TIL maps, aggressiveness maps, bone-marrow ROI/cell annotations) converted from CSV/XML/PNG/TXT/HDF5 sources into standard DICOM objects (SEG, PM, SR TID1500, MBSA) and published in the NCI Imaging Data Commons, each with a Zenodo DOI and IDC portal entry.
How can I access or download the converted datasets?
Via the IDC portal (analysis-result or original collections), Zenodo records listed in the Data Availability section, cloud buckets on AWS and Google Cloud, DICOMweb against a Google Healthcare DICOM Store, BigQuery metadata queries, or the idc-index Python library; Slim is the reference web viewer.
Which DICOM object type should I use for my pathology annotations?
The paper's guidance (Figure 5): SEG or PM for raster masks/maps (labelmap SEG for many discrete segments), SR TID1500 for few heterogeneous regions with rich evaluations, and MBSA for large numbers of homogeneous vector ROIs; use tiled-full encoding and image-relative coordinates as pragmatic defaults.
Is the conversion code and tooling open source?
Yes — conversion code is on GitHub (ImagingDataCommons/idc-sm-annotations-conversion and conversion_mirax_dicom) with versioned Zenodo archives, and the work added substantial capabilities to the open-source highdicom library (v0.21.0–0.28.0) and Slim viewer.
About this review
This is an AI-generated preliminary review by Evolve.Pub, produced from the openly available text of the paper. It is not peer review and does not represent the authors' views. The canonical record of the paper is at the publisher link above.
- Figures 1, 2, and 5 are available only as captions; their visual content (schematics, viewer screenshots, flowchart) could not be analyzed.
- Tables 1 and 2 were extracted with garbled layout; all numbers used were cross-checked against the prose, but the BoneMarrow 2452-slide figure could conceivably be an extraction artifact.
- Review is based on the arXiv preprint text; external resources (IDC portal, Zenodo, dashboard, notebooks) were not verified.
- The paper notes it may be truncated relative to any final published version; supplementary details beyond Table S1 were not present.
Go deeper on this paper
Wrote a paper? Stress-test it before reviewers do.
Upload your manuscript for the same structured scrutiny — reasoning, statistics, evidence, figures, and reproducibility.