Lumbar Spine MRI Dataset
Description
This data set contains anonymised clinical MRI study, or a set of scans, of 515 patients with symptomatic back pains. Each patient data can have one or more MRI studies associated with it. Each study contains slices, i.e., individual images taken from either sagittal or axial view, of the lowest three vertebrae and the lowest three IVDs. The axial view slices are mainly taken from the last three IVDs – including the one between the last vertebrae and the sacrum. The orientation of the slices of the last IVD are made to follow the spine curve whereas those of the other IVDs are usually made in blocks – i.e., parallel to each other. There are between four to five slices per IVD and they begin from the top of the IVD towards its bottom. Many of the top and bottom slices cut through the vertebrae leaving between one to three slices that cut the IVD cleanly and show purely the image of that IVD. In most cases, the total number of slices in axial view ranges from 12 to 15. However, in some cases, there may be up to 20 slices because the study contains slices of more than last three vertebrae. The scans in sagittal view also vary but all contain at least the last seven vertebrae and the sacrum. While the number of vertebrae varies, each scan always includes the first two sacral links. There are a total 48,345 MRI slices in our dataset. The majority of the slices have an image resolution of 320x320 pixels, however, there are slices from three studies with 320x310 pixel resolution. The pixels in all slices have 12-bit per pixel precision which is higher than the standard 8-bit greyscale images. Specifically for all axial-view slices, the slice thickness are uniformly 4 mm with centre-to-centre distance between adjacent slices to be 4.4 mm. The horizontal and vertical pixel spacing is 0.6875 mm uniformly across all axial-view slices. The majority of the MRI studies were taken with the patient in Head-First-Supine position with the rests were taken with the patient in in Feet-First-Supine position. Each study can last between 15 to 45 minutes and a patient may have one or more study associated with them taken at a different time or a few days apart. You can download and read the research papers detailing our methodology on boundary delineation for lumbar spinal stenosis detection using the URLs provided in the Related Links at the end of this page. You can also check out other dataset and source code related to this program from that section. We kindly request you to cite our papers when using our data or program in your research.
Files
Steps to reproduce
The data is anonymised and is provided as is. A DICOM viewer maybe necessary to view this dataset.