GTU Lumbar MRI Dataset
Lumbar MRI Dataset of the Computer Vision Lab, Gebze Technical University
Mid-sagittal T1- and T2-weighted lumbar spine MR slices of 80 subjects (64 with disc or vertebra pathologies), with expert-marked centre points of the L1 to L5 vertebrae and the T12-L1 to L5-S1 discs for localization and labelling.
Overview
The GTU Lumbar MRI Dataset comes from the Computer Vision Lab at Gebze Technical University and Istanbul Medeniyet University. It was released with a 2013 IEEE Transactions on Biomedical Engineering paper by Oktay and Akgul on finding and labelling lumbar vertebrae and intervertebral discs in sagittal MR slices, so that other methods could be compared on the same data. It is a small benchmark for 2D spine landmark localization.
Composition
The release holds 80 anonymous subjects. For each subject there is one T1-weighted and one T2-weighted mid-sagittal
slice, stored as MATLAB .mat files, 160 images in total. Each slice shows the five lumbar vertebrae and six lumbar
discs, plus parts of the thoracic spine and sacrum. According to the paper, 16 subjects are free of pathology and 64
have disc or vertebra problems. Across the cohort, 158 lumbar intervertebral structures show findings such as
herniation, degeneration or scoliosis, and 31 vertebra abnormalities such as lumbarization or nodular lesions are
present. The zip download is about 44 MB.
Acquisition
The paper describes the source as a clinical MR dataset with three protocols per subject: sagittal T1-weighted, sagittal T2-weighted and axial T2-weighted. Only the two mid-sagittal slices are part of the release. Images are 512 by 512 pixels with 0.625 mm pixel spacing and 4 mm between sagittal slices. Scanner, field strength and site are not stated.
Annotations
One expert marked the approximate centre of each structure. The expert folder gives, per subject, eleven points from
the T12-L1 disc down to the L5-S1 disc, alternating discs and vertebrae L1 to L5. The our_results folder holds the
points predicted by the authors' method, and a MATLAB script overlays both on the image. The paper also mentions expert
contours of each structure, but the dataset page lists only the centre points.
Known limitations
- Single 2D slices only, no volumes, and no age, sex or diagnosis per subject.
- No license or terms of use are published, so commercial use, model training and redistribution are not settled.
- Files are in MATLAB format and need conversion for most other tools.
Cohort
Aggregate numbers from the sources below. Bars are relative to the 80 subjects.
Contrast combinations
How many subjects have exactly each set of contrasts.
| T1w | T2w | Subjects with exactly this set |
|---|---|---|
80 |
Contrast / sequence
subjects, values can overlap
- T1-weighted 80 100%
- T2-weighted 80 100%
Condition
subjects, values can overlap
- Healthy control 16 20%
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
No published license or data use terms
The data holder has published no license and no data use terms. Any use, sharing or commercial right has to be agreed with the data holder, and nothing can be assumed to be allowed. Default copyright and data protection law still apply.
The dataset page links a public zip download and asks users to cite two papers, but states no license or terms of use. The paper only says the data can be downloaded. Use beyond citation has to be agreed with the authors.
What you can do
- Commercial use Not stated
- Train ML models Not stated
- Create derived data Not stated
- Publish results Not stated
What you can share
- Share the data Not stated
- Share derived data Not stated
- Share trained models Not stated
What you must do
- Cite or credit Not stated
- Share alike Not stated
- Sign an agreement Not stated
- Ethics approval Not stated
- Manuscript review Not stated
- Release code Not stated
- Return results Not stated
- Delete after use Not stated
Limits
- No re-identification Not stated
- Location limits Not stated
Citation
Oktay AB, Akgul YS. Simultaneous Localization of Lumbar Vertebrae and Intervertebral Discs With SVM-Based MRF. IEEE Transactions on Biomedical Engineering 60(9), 2375-2383 (2013). doi:10.1109/TBME.2013.2256460
@article{oktay2013,
title={Simultaneous Localization of Lumbar Vertebrae and Intervertebral Discs With SVM-Based MRF},
author={Oktay, Ayse Betul and Akgul, Yusuf Sinan},
journal={IEEE Transactions on Biomedical Engineering},
volume={60},
number={9},
pages={2375--2383},
year={2013},
doi={10.1109/TBME.2013.2256460}
} All numbers
Every number on this page, as stored in stats.csv, with its source.
| Measure | Breakdown | Value | Source |
|---|---|---|---|
| Subjects | total | 80 | oktay2013 Section V |
| Subjects | condition=healthy pathology-free; the other 64 subjects have disc or vertebra pathologies | 16 | oktay2013 Section V |
| Subjects | contrast=T1w one T1-weighted mid-sagittal image per subject (s_T1.mat) | 80 | gtu-lumbar-mri-page Data Description |
| Subjects | contrast=T2w one T2-weighted mid-sagittal image per subject (s_T2.mat) | 80 | gtu-lumbar-mri-page Data Description |
| Subjects | contrast_set=T1w+T2w | 80 | gtu-lumbar-mri-page Data Description |
| Images | total 80 subjects times one T1 and one T2 mid-sagittal image, summed from the file naming | 160 | gtu-lumbar-mri-page Data Description |
| Images | contrast=T1w | 80 | gtu-lumbar-mri-page Data Description |
| Images | contrast=T2w | 80 | gtu-lumbar-mri-page Data Description |
Sources
The keys used in the table above.
- gtu-lumbar-mri-page GTU VisLab Lumbar MRI Dataset page website
- oktay2013 Oktay and Akgul 2013, IEEE Transactions on Biomedical Engineering (author copy) paper