license: cc-by-4.0
task_categories:
- image-segmentation
tags:
- medical
- mri
- ct
- pet
- cervix
- cervical-cancer
- tumor-segmentation
- longitudinal
- chemoradiation
- rtstruct
- dicom
- tcia
pretty_name: TCIA Cervical Cancer (CC-Tumor-Heterogeneity)
size_categories:
- n<1K
TCIA_CervicalCancer (CC-Tumor-Heterogeneity)
Longitudinal multi-parametric MRI + 18F-FDG PET/CT collection of locally advanced cervical cancer patients undergoing chemoradiation, with radiation-oncologist tumor contours (RTSTRUCT) defined on T2-weighted MRI as the gold standard. Each patient was imaged at three time points spanning the chemoradiation course.
Dataset Details
| Field | Value |
|---|---|
| Modalities | MR (T1W, T2W, DCE, DWI), 18F-FDG PET, CT |
| Body part | Cervix / pelvis |
| Task | 3D tumor segmentation on T2W MRI (longitudinal) |
| Patients | 23 (CCTH-A01 … CCTH-A12, CCTH-B01 … CCTH-B11) |
| Time points | 3 per patient — Baseline (0 Gy) / Early ( |
| Studies | 171 |
| Series | 821 |
| Images | 131,556 DICOM slices |
| MR series | 523 (15.91 GB) |
| CT series | 65 (11.48 GB) |
| PET series | 68 (0.99 GB) |
| RTSTRUCT files | 68 (≈ 23 patients × 3 time points; CCTH-A06 has 2) |
| REG (registration) files | 97 |
| Format | DICOM (images) + DICOM RTSTRUCT (contours) + DICOM REG (registrations) |
| License | CC BY 4.0 |
Segmentation
Tumor contours are stored as DICOM RTSTRUCT files in segmentations/.
Each RTSTRUCT references a single MR T2W sagittal series via
ReferencedFrameOfReferenceSequence → RTReferencedStudySequence → RTReferencedSeriesSequence.SeriesInstanceUID — that referenced series is the
image volume the contour is defined on. ROI names follow Ut-MRT2-Sag-{1|2|3}
where the trailing index matches the time point (1 = Baseline, 2 = Early,
3 = Mid-treatment), and StructureSetName carries Timepoint{1|2|3}.
The collection paper describes contours as tumor volume only — no OARs and no lymph nodes, drawn by the study radiation oncologists with T2W MRI chosen for its highest soft-tissue contrast for cervical tumor delineation.
Structure
images_mr/<PatientID>/<StudyInstanceUID>/<SeriesInstanceUID>/*.dcm # MR (T1W, T2W, DCE, DWI variants)
images_ct/<PatientID>/<StudyInstanceUID>/<SeriesInstanceUID>/*.dcm # CT
images_pt/<PatientID>/<StudyInstanceUID>/<SeriesInstanceUID>/*.dcm # PET
segmentations/<PatientID>/<StudyInstanceUID>/<SeriesInstanceUID>/*.dcm # RTSTRUCT (one per time point)
registrations/<PatientID>/<StudyInstanceUID>/<SeriesInstanceUID>/*.dcm # DICOM REG (PET/CT ↔ MR coregistration)
series_to_patient.json # series-level index
PatientID ranges over CCTH-A01..A12 and CCTH-B01..B11. StudyDescription
in each RTSTRUCT carries MR1 / MR2 / MR3, identifying the time point.
CCTH-B11 has no CT series; CCTH-A06 has 2 RTSTRUCTs (one time point missing).
Source
- TCIA collection: https://www.cancerimagingarchive.net/collection/cc-tumor-heterogeneity/
- DOI:
10.7937/ERZ5-QZ59 - License URL: https://creativecommons.org/licenses/by/4.0/
- Released: 2023-01-20 (now fully public, no registration required)
Citation
@misc{mayr2023ccth,
author = {Mayr, N. and Yuh, W. T. C. and Bowen, S. and Harkenrider, M. and
Knopp, M. V. and Lee, E. Y.-P. and Leung, E. and Lo, S. S. and
Small Jr., W. and Wolfson, A. H.},
title = {Cervical Cancer -- Tumor Heterogeneity: Serial Functional and
Molecular Imaging Across the Radiation Therapy Course in
Advanced Cervical Cancer (Version 1) [Data set]},
year = {2023},
publisher = {The Cancer Imaging Archive},
doi = {10.7937/ERZ5-QZ59}
}
@article{bowen2017radiomic,
author = {Bowen, S. R. and Yuh, W. T. C. and Hippe, D. S. and Wu, W. and
Partridge, S. C. and Elias, S. and Jia, G. and Huang, Z. and
Sandison, G. A. and Nelson, D. and Knopp, M. V. and Lo, S. S. and
Kinahan, P. E. and Mayr, N. A.},
title = {Tumor radiomic heterogeneity: Multiparametric functional imaging
to characterize variability and predict response following
cervical cancer radiation therapy},
journal = {Journal of Magnetic Resonance Imaging},
volume = {47},
number = {5},
pages = {1388--1396},
year = {2017},
doi = {10.1002/jmri.25874}
}