Skip to content
Open Imaging Index
MRI Spine

Multi-scanner lumbar spine MRI segmentation database

Multi-scanner and multi-modal lumbar vertebral body and intervertebral disc segmentation database

Clinical lumbar spine MRI from three scanner makers with manual masks of vertebral bodies and discs, built to test segmentation that copes with different scanners and sequences. 34 patients across 55 sessions from Munich.

Go to dataset Edit Report

Overview

This database holds routine lumbar spine MRI from patients of the Klinikum rechts der Isar in Munich, with manual masks of the lumbar vertebral bodies and intervertebral discs. It was assembled by radiologists at the Technical University of Munich together with Eindhoven University of Technology. Its purpose is to train and test segmentation methods that have to cope with images from different scanners and pulse sequences, rather than from one tuned protocol.

Composition

There are 34 adult patients (20 female, 14 male) aged 30 to 88 years. They were scanned for low back pain or radiculopathy, follow-up after resection of a spinal tumour, known cancer with suspected spinal metastases, spondylodiscitis or other inflammation, and in one case a sacral fracture. Twenty-one patients have a second session on another scanner model or with another protocol, between two weeks and five years after the first, which gives 55 sessions. Every session has a non-contrast sagittal T1-weighted series. Depending on the session there are also contrast-enhanced T1-weighted, fat-saturated contrast T1-weighted, T2-weighted, STIR and T2-weighted Dixon series; for Dixon the water, fat and in-phase images are separate files. A spreadsheet lists sex, age, indication, scanner and sequences per patient.

Acquisition

Scans date from August 2014 to October 2019, partly from the hospital's own scanners and partly transferred from other institutions. About 71% of sessions come from Philips systems (Achieva, Ingenia, Elition), about 27% from seven Siemens models and one from a GE Signa. Each patient has at least one Philips session with a reference sagittal T1 and a T2 Dixon turbo spin echo protocol at roughly 0.8 x 1 x 3 mm voxels. All images are NIfTI files.

Annotations

A medical doctor outlined the vertebral bodies L1 to L5 and the discs L1/2 to L4/5 by hand in the sagittal plane of the non-contrast T1 images in MITK, under the supervision of a radiologist with eleven years of experience. Posterior elements are excluded. The other series of each patient were registered to that T1 image and resampled into its space, so every volume comes with nine binary masks, one per structure.

Known limitations

  • Small cohort from one hospital, with mixed clinical indications.
  • Field strength is not reported.
  • Labels exist only in the T1 space; other series were resampled with linear interpolation, which smooths them.
  • One annotator; no inter-rater agreement is given.

Cohort

Aggregate numbers from the sources below. Bars are relative to the 34 subjects.

Sex

  • Female 20 59%
  • Male 14 41%

Table 1 in Al Khalil et al. 2022, Scientific Data 9, 97

Contrast / sequence

Groups can overlap

  • T1-weighted 34 100%
  • T2-weighted 34 100%
  • T1-weighted, contrast enhanced 22 65%
  • STIR 7 21%

Tables 2 and 3 in Al Khalil et al. 2022, Scientific Data 9, 97

Scanner vendor

  • Philips 34 100%
  • Siemens Healthineers 15 44%
  • GE HealthCare 1 3%

Methods: Patient cohort, Tables 2 and 3 in Al Khalil et al. 2022, Scientific Data 9, 97

Age

mean 60.4 ± 15.2, range 30 to 88.1

0
2
4
6
8
8
30-3940-4950-5960-6970-7980-89

Table 1, Methods: Patient cohort in Al Khalil et al. 2022, Scientific Data 9, 97

Condition

Groups can overlap

  • Low back pain 14 41%
  • Discitis 4 12%
  • Bone fracture 1 3%

Table 1 in Al Khalil et al. 2022, Scientific Data 9, 97

Contrast combinations

How many subjects have exactly each set of contrasts.

T1wT1w_ceT2wSTIRSubjects with exactly this set
18
9
4
3

Tables 2 and 3 in Al Khalil et al. 2022, Scientific Data 9, 97

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

Access
Open download

Download without an account

Access page

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.

Original license text Version read: 4.0 Checked 2026-10-07

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

Citation

Al Khalil Y, Becherucci EA, Kirschke JS, Karampinos DC, Breeuwer M, Baum T, Sollmann N. Multi-scanner and multi-modal lumbar vertebral body and intervertebral disc segmentation database. Sci Data 9, 97 (2022). https://doi.org/10.1038/s41597-022-01222-8

@article{alkhalil2022,
  title = {Multi-scanner and multi-modal lumbar vertebral body and intervertebral disc segmentation database},
  author = {Al Khalil, Yasmina and Becherucci, Edoardo A. and Kirschke, Jan S. and Karampinos, Dimitrios C. and Breeuwer, Marcel and Baum, Thomas and Sollmann, Nico},
  journal = {Scientific Data},
  volume = {9},
  pages = {97},
  year = {2022},
  doi = {10.1038/s41597-022-01222-8}
}

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.