UT-EndoMRI
UTHealth Endometriosis MRI Dataset
Pelvic MRI of women with suspected or confirmed endometriosis, released to develop and test automatic segmentation of the ovaries, uterus and endometriomas, with manual labels from up to three raters for studying interrater agreement.
Overview
UT-EndoMRI is a set of pelvic MRI scans with manual organ and lesion labels from patients examined for endometriosis at two hospital systems in Houston, Texas. A team at the McWilliams School of Biomedical Informatics (UTHealth Houston) assembled it to measure how well raters agree on ovary outlines and to build an automatic ovary segmentation method, RAovSeg, with nnU-Net as the baseline. The images are deposited on Zenodo as NIfTI files.
Composition
The release has two parts. Dataset 1 (folder D1_MHS) holds 51 patients scanned before 2022 within the Memorial Hermann Hospital System; all were suspected of endometriosis, and eight were not diagnosed with it. Dataset 2 (folder D2_TCPW) holds 81 endometriosis patients scanned in 2022 at Texas Children's Hospital Pavilion for Women; the Zenodo description gives 82. Dataset 1 contains T2-weighted and fat-suppressed T1-weighted scans, Dataset 2 contains T1-weighted and T2-weighted scans with and without fat suppression. Not every patient has every sequence.
Acquisition
Dataset 1 comes from 15 sites and nine scanner models by GE, Philips and Siemens at 1.5 T and 3 T. A file on Zenodo lists site and scanner per subject, with the entry unknown for 18 subjects. Dataset 2 was acquired at one site on a Philips Ingenia 1.5 T. DICOM images were converted to NIfTI for de-identification, and the sequences of each subject are registered to each other.
Annotations
In Dataset 1, uterus, ovaries and endometriomas were outlined in 3D Slicer by up to three raters following a guideline from an abdominal radiologist, who reviewed the final labels; 11 subjects have three raters, 22 have two and 18 have one. The Zenodo files also include cyst and cul-de-sac labels. In Dataset 2, one obstetrics and gynecology assistant, supervised by a gynecologist, outlined uterus, ovaries, cysts and endometriomas on the fat-suppressed T2-weighted scans.
Known limitations
Labels are missing where a structure was absent or not visible, and only 12 patients in Dataset 2 have endometriomas. Interrater agreement on the ovaries is moderate. The paper and the Zenodo record disagree on the Dataset 2 subject count and on which subjects formed the RAovSeg training and test sets. Table 1 of the paper lists plain T1-weighted scans for Dataset 1, while its text names only fat-suppressed T1-weighted scans. Zenodo lists CC BY 4.0, while the record's user agreement limits use to non-commercial research.
Cohort
Aggregate numbers from the sources below. Bars are relative to the 132 subjects.
Condition
Groups can overlap
- Endometriosis 124 94%
Methods, Image acquisition in Liang et al. 2025, Scientific Data
Contrast / sequence
ScansGroups can overlap
- T1-weighted 228 57%
- T2-weighted 172 43%
Table 1 in Liang et al. 2025, Scientific Data
License and access
Our reading of the license, not legal advice. Before you use the data, read the original license and confirm that your use is allowed. We take no responsibility for how you use a dataset. Full disclaimer
Download without an account
Different parts of the data carry different licenses. The summary on the right shows the most restrictive answer per rule.
Creative Commons Attribution 4.0 International
Use, share and adapt the data for any purpose, including commercial use, as long as you credit the creators.
What you can do
- Yes
- Yes
- Yes
- Yes
- Yes
What you can share
- Yes
- Yes
- Yes
What you must do
- Yes
- Share alike No
- No
- No
- Manuscript review No
- Release code No
- Return results No
- Delete after use No
Limits
- No
- Location limits No
UT-EndoMRI User Agreement (Zenodo record description)
A short statement in the Zenodo record that limits use of the dataset to non-commercial scientific research and requires publications to cite the data descriptor by Liang et al. (Scientific Data 2025). It is silent on redistribution, derived data and trained models. The same record lists CC BY 4.0 as its license.
The record carries CC BY 4.0 in its license field, while its description limits use to non-commercial scientific research. Both are listed so the stricter terms are shown.
What you can do
- No
- Not stated
- Not stated
- Create derived data Not stated
- Conditional
What you can share
- Share the data Not stated
- Share derived data Not stated
- Share trained models Not stated
What you must do
- Yes
- Share alike No
- Sign an agreement No
- Ethics approval No
- Manuscript review No
- Release code No
- Return results No
- Delete after use No
Limits
- No
- Location limits No
Citation
Liang X, Alpuing Radilla LA, Khalaj K, Dawoodally H, Mokashi C, Guan X, Roberts KE, Sheth SA, Tammisetti VS, Giancardo L. A Multi-Modal Pelvic MRI Dataset for Deep Learning-Based Pelvic Organ Segmentation in Endometriosis. Scientific Data 12, 1292 (2025). https://doi.org/10.1038/s41597-025-05623-3
@article{liang2025,
title={A Multi-Modal Pelvic MRI Dataset for Deep Learning-Based Pelvic Organ Segmentation in Endometriosis},
author={Liang, Xiaomin and Alpuing Radilla, Linda A. and Khalaj, Kamand and Dawoodally, Haaniya and Mokashi, Chinmay and Guan, Xiaoming and Roberts, Kirk E. and Sheth, Sunil A. and Tammisetti, Varaha S. and Giancardo, Luca},
journal={Scientific Data},
volume={12},
pages={1292},
year={2025},
doi={10.1038/s41597-025-05623-3}
} Sources
Every number on this page comes from one of these documents. Each chart names the table or page it is taken from. The raw numbers are in stats.csv.