Lumbar Spine MRI Dataset
Published: 3 April 2019
Version: 2
Contributors
- Sud Sudirman
- Ala Al Kafri
- Friska Natalia
- Hira Meidia
- Nunik Afriliana
- Wasfi Al-Rashdan
- Mohammad Bashtawi
- Mohammed Al-Jumaily
Description
This dataset contains anonymised clinical MRI studies (or sets of scans) of 515 patients with symptomatic back pain. Each patient may have one or more MRI studies associated with them.
Each study contains slices—individual images—captured from either sagittal or axial views of the lowest three vertebrae and the lowest three intervertebral discs (IVDs). The axial view slices focus on the last three IVDs, including the one between the last vertebra and the sacrum.
The slices of the last IVD follow the spine's curvature, while others are acquired in blocks (i.e., parallel slices). Typically, there are 4–5 slices per IVD, moving from top to bottom. Some top and bottom slices cut into the vertebrae, leaving 1–3 clean IVD-only slices.
Most axial view studies contain 12 to 15 slices, though some include up to 20 slices if more vertebrae are scanned. Sagittal scans include at least the last seven vertebrae and the sacrum, always covering the first two sacral links.
The dataset comprises 48,345 MRI slices. Most images have a resolution of 320×320 pixels, with three studies using 320×310. All slices have 12-bit pixel precision, providing higher detail than standard 8-bit grayscale images.
Axial View Slice Specs:
- Slice thickness: 4 mm
- Slice spacing (center-to-center): 4.4 mm
- Pixel spacing: 0.6875 mm (both directions)
The majority of scans were taken in the Head-First-Supine position, with some in Feet-First-Supine. Each study lasts between 15–45 minutes. Multiple studies may exist for one patient, conducted on different days.
You can download and read research papers describing the boundary delineation method for lumbar spinal stenosis detection using the links provided in the Related Links section of the dataset page.
We kindly request that you cite our papers when using this dataset or any related software in your research.