Lumbar Spine VCF MRI (HCFMRP)
Lumbar Spine Vertebral Compression Fractures (VCFs) Dataset: MRI T1-Weighted Images for Benign and Malignant Classification
Sagittal T1-weighted lumbar spine MRI of adults with vertebral compression fractures, collected to tell osteoporotic from cancer-related fractures with 3D radiomics, with semi-automatic segmentations of each fractured vertebral body.
Overview
A retrospective single-centre cohort of adults with lumbar vertebral compression fractures, assembled at the University Hospital of the Ribeirão Preto Medical School (HCFMRP), University of São Paulo, Brazil. The study behind it trained an artificial neural network on three-dimensional radiomic features to separate benign fractures caused by osteoporosis from malignant fractures caused by tumour infiltration of the vertebra. The images, segmentations and a per-patient table were released on Zenodo in August 2024 under CC BY 4.0.
Composition
91 patients (55 women, 36 men) diagnosed between 2010 and 2019, mean age about 64 years. 47 have benign and 44 have malignant fractures. Together they have 146 fractured lumbar vertebrae. Each patient contributes one examination, the earliest one in the hospital archive. The paper held out the last 15 benign and last 15 malignant patients retrieved from the archive as a test set, with 61 patients for training.
Excluded were patients treated with chemotherapy, radiotherapy or surgery before the scan, traumatic and old fractures, patients under 18, and malignant cases without a biopsy.
Acquisition
Routine clinical sagittal T1-weighted lumbar spine MRI on Philips Achieva scanners at 1.5 T and 3 T. Slice thickness is 3 to 5 mm with 12 to 15 slices per series. Images are shared as anonymized DICOM and as NRRD volumes rescaled to 256 grey levels after histogram equalization.
Annotations
Each fractured vertebral body was segmented in 3D with threshold and erosion tools in 3D Slicer by a trained investigator and reviewed by a musculoskeletal radiologist with 20 years of experience. Labels come from spine biopsy for malignant fractures and from DEXA, imaging reread and two years of follow-up for benign ones. The spreadsheet lists class, sex, age and the segmented vertebral levels per patient.
Known limitations
- Single centre, single vendor and one contrast only, so no T2-weighted, STIR or contrast-enhanced series.
- The malignant group is not broken down by tumour type in the public description.
- The Zenodo description gives a mean age of 64.24 years and the paper 64.75 years for the same patients.
- Field strength per patient is not reported in the description or paper text.
Cohort
Aggregate numbers from the sources below. Bars are relative to the 91 subjects.
Condition
Groups can overlap
- Bone fracture 91 100%
- Osteoporosis 47 52%
- Cancer 44 48%
Split
- Training 61 67%
- Test 30 33%
Material and Methods: Study Design in Chiari-Correia et al. 2023, Journal of Digital Imaging
Contrast / sequence
Groups can overlap
- T1-weighted 91 100%
Scanner vendor
- Philips 91 100%
Material and Methods: Study Design in Chiari-Correia et al. 2023, Journal of Digital Imaging
Contrast combinations
How many subjects have exactly each set of contrasts.
| T1w | Subjects with exactly this set |
|---|---|
91 |
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
Direct download from Zenodo without an account: one RAR archive (about 1.3 GB) with the anonymized DICOM exams, intensity-requantized NRRD volumes, seg.nrrd segmentations and a per-patient spreadsheet.
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.
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
Chiari-Correia NS, Nogueira-Barbosa MH, Chiari-Correia RD, Azevedo-Marques PM. A 3D Radiomics-Based Artificial Neural Network Model for Benign Versus Malignant Vertebral Compression Fracture Classification in MRI. J Digit Imaging. 2023;36(4):1565-1577. doi:10.1007/s10278-023-00847-4
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.