Skip to content
Open Imaging Index
MRI Intervertebral disc

IVDM3Seg

IVDM3Seg: Automatic Intervertebral Disc Localization and Segmentation from 3D Multi-modality MR (M3) Images (MICCAI 2018 challenge, SpineWeb Dataset 14)

1.5 T Siemens Dixon MRI of the lower spine from 12 subjects scanned at two stages of a prolonged bed rest study (24 scan sets, 96 volumes), with manual masks of at least 7 intervertebral discs. Training part of 8 subjects shared after a signed agreement.

Go to dataset Edit Report

Overview

IVDM3Seg is the dataset of the MICCAI 2018 challenge on automatic localization and segmentation of intervertebral discs from 3D multi-modality MR images, held on 16 September 2018 in Granada together with the workshop on Computational Methods and Clinical Applications for Spine Imaging. Guoyan Zheng (University of Bern), Daniel Belavy (Deakin University) and Shuo Li (University of Western Ontario) organized it, and SpineWeb lists it as Dataset 14. The challenge paper by Zeng et al. compares 8 submitted methods, run as Docker containers by the organizers.

Composition

12 subjects, each scanned at two stages of a study on the effect of prolonged bed rest (a spaceflight simulation) on the lumbar discs. This gives 24 multi-modality data sets. Each data set has four aligned 3D volumes, for 96 volumes in total, and covers at least 7 discs of the lower spine. The agreement form places the discs between T11 and L5. Images and masks are NIfTI files. The training part, released on 15 March 2018, holds 16 data sets from 8 subjects with their masks. The remaining data served as the hidden test set for the on-site competition. No age, sex or health status is given.

Acquisition

All scans were made on a 1.5 T Siemens scanner with a Dixon protocol, which reconstructs in-phase, opposed-phase, fat and water volumes from one acquisition. Sequence parameters, voxel size and site are not stated on the challenge pages.

Annotations

Every disc has a manual reference segmentation, stored as one binary mask volume per data set. The challenge asks for the centre of each of the 7 discs (localization) and a disc versus background labelling (segmentation), ranked by Dice, average surface distance and localization distance. The pages do not say who drew the masks.

Known limitations

  • Small: 12 subjects from one scanner type and one protocol, and only 8 subjects are released.
  • Test masks are held by the organizers and are not distributed.
  • No demographics, acquisition parameters or annotation protocol are published on the challenge pages.
  • The SpineWeb host no longer resolved when this entry was checked on 2026-10-09; its Dataset 14 entry was read from an Internet Archive copy of June 2024.

Cohort

Aggregate numbers from the sources below. Bars are relative to the 12 subjects.

Contrast combinations

How many subjects have exactly each set of contrasts.

dixonSubjects with exactly this set
12

Contrast / sequence

subjects, values can overlap

  • Dixon fat-water 12 100%

Anatomy

subjects

  • Intervertebral disc 12 100%

Scanner vendor

subjects

  • Siemens Healthineers 12 100%

Field strength

subjects

  • 1.5 T 12 100%

Split

subjects

  • Training 8 67%

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
Signed agreement

Sign a data use agreement, often reviewed by the provider

Access page
For Training data (16 scan sets of 8 subjects)

Agreement on using the Intervertebral Disc Localization and Segmentation Multi-modality MRI Spine Image Database

A one-page registration form addressed to the Institute for Surgical Technology and Biomechanics, University of Bern, that the user signs to receive a download link. It forbids passing on the link or the data, requires citing a 2014 MLMI paper in every publication and sending copies of those publications to the provider. It says nothing about purpose or commercial use.

The test data were kept by the organizers for the on-site competition and are not released. The SpineWeb entry for Dataset 14 says to contact Dr. Zheng to obtain the data.

Original license text Version read: Registration form linked from the IVDM3Seg Data page, undated, read 2026-10-09 Checked 2026-10-09

What you can do

  • Not stated
  • Train ML models Not stated
  • Create derived data Not stated
  • Conditional

What you can share

  • No
  • Share derived data Not stated
  • Share trained models Not stated

What you must do

  • Yes
  • Share alike No
  • Yes
  • No
  • No
  • Release code No
  • Yes
  • Delete after use No

Limits

  • No
  • Location limits No

Citation

Chen C, Belavy D, Zheng G. 3D Intervertebral Disc Localization and Segmentation from MR Images by Data-Driven Regression and Classification. MLMI 2014, LNCS 8679, pp. 50-58. doi:10.1007/978-3-319-10581-9_7

All numbers

Every number on this page, as stored in stats.csv, with its source.

MeasureBreakdownValueSource
Subjectstotal
each subject scanned at two stages of a prolonged bed rest study
12ivdm3seg-data-page
Data Description
Studiestotal
24 multi-modality data sets, one per subject per stage
24ivdm3seg-data-page
Data Description
Scanstotal
four aligned Dixon volumes per data set: in-phase, opposed-phase, fat and water
96ivdm3seg-data-page
Data Description
Subjectscontrast=dixon 12ivdm3seg-data-page
Data Description
Subjectscontrast_set=dixon
the page lists only the four Dixon volumes per data set
12ivdm3seg-data-page
Data Description
Subjectsanatomy=intervertebral_disc
at least 7 lower-spine discs per data set with manual binary masks
12ivdm3seg-data-page
Data Description
Subjectsfield_strength=1.5 12ivdm3seg-data-page
Data Description
Subjectsvendor=siemens 12ivdm3seg-data-page
Data Description
Subjectssplit=train 8ivdm3seg-data-page
Training Data
Studiessplit=train 16ivdm3seg-data-page
Training Data

Sources

The keys used in the table above.