Skip to content
Open Imaging Index
CT MRI NeckSpine

MMCSD

Multi-modal and Multi-view Cervical Spondylosis Imaging Dataset

Preoperative cervical spine MRI and CT of patients with cervical spondylosis and neck pain who had anterior fusion surgery, paired with neck pain scores before and one year after surgery to predict postoperative pain. 250 patients from one hospital in China.

Go to dataset Edit Report

Overview

MMCSD pairs preoperative cervical spine MRI and CT with neck pain outcomes for patients operated on for cervical spondylosis at the First Medical Center of the Chinese PLA General Hospital. The authors built it so that deep learning models can be trained to predict whether neck pain persists after surgery, and they suggest it for self-supervised pretraining and MRI-to-CT synthesis as well. A baseline 3D DenseNet121 trained per sequence reached an accuracy of 0.74 on sagittal T2-weighted images.

Composition

The cohort is 250 patients treated between January 2019 and December 2022. Inclusion required a diagnosis of cervical spondylotic myelopathy (108) or radiculopathy (122), with 20 of mixed type, anterior cervical discectomy and fusion or corpectomy and fusion, and a neck pain score of at least 3 on the visual analogue scale. Mean age is 54.8 years (SD 10.7). Each patient has five image volumes: axial CT in bone window and in soft tissue window, sagittal T1-weighted and T2-weighted MRI, and axial T2-weighted MRI. The Synapse page counts 3,289 image slices in total. A spreadsheet holds sex, age, diagnosis, surgical approach and levels, and the pain score before surgery and one year after. A fixed split assigns 200 patients to training and 50 to testing.

Acquisition

MRI came from 13 scanners by United Imaging, GE, Philips, Siemens and Alltech, at 3 T for 143 patients and 1.5 T for 107. CT came from five scanners by Philips, GE, United Imaging and NMS, most often a Philips iCT 256. The paper lists sequence and CT parameters per vendor. Volumes were cropped to the cervical spine: MRI keeps 8 to 12 slices per sequence, and CT was subsampled at 1:10 to 1:20 to 10 to 20 slices. Files are NIfTI with identifying header fields removed.

Annotations

There are no segmentations. The label is the postoperative pain outcome; the baseline treats any neck pain score above 0 at one year as persistent pain.

Known limitations

  • One hospital and only surgical patients with neck pain, so the cohort does not represent cervical spondylosis in general.
  • CT is heavily subsampled and MRI is cropped, so the volumes are not full diagnostic series.
  • The sex counts in the paper's text and Table 5 are swapped (134 female or 134 male), so they are left out here.

Cohort

Aggregate numbers from the sources below. Bars are relative to the 250 subjects.

Age

mean 54.8 ± 10.7

No age bins reported.

Table 5 in Yu et al. 2025, Scientific Data 12, 1080

Contrast / sequence

Groups can overlap

  • T1-weighted 250 100%
  • T2-weighted 250 100%

Data Records in Yu et al. 2025, Scientific Data 12, 1080

Field strength

  • 3 T 143 57%
  • 1.5 T 107 43%

Table 1 in Yu et al. 2025, Scientific Data 12, 1080

Scanner vendor by modality

Reported cross table. A dot marks a cell the source does not give.

MRICT
Philips42163
United Imaging10538
GE HealthCare7241
Siemens Healthineers28·
Other38

Table 1, Table 2 in Yu et al. 2025, Scientific Data 12, 1080

Modality

  • MRI 250 100%
  • CT 250 100%

Data Records in Yu et al. 2025, Scientific Data 12, 1080

Condition

Groups can overlap

  • Cervical spondylosis 250 100%
  • Cervical myelopathy 108 43%

Methods: Subject characteristics, Table 5 in Yu et al. 2025, Scientific Data 12, 1080

Split

  • Training 200 80%
  • Test 50 20%

Table 5 in Yu et al. 2025, Scientific Data 12, 1080

Contrast combinations

How many subjects have exactly each set of contrasts.

T1wT2wSubjects with exactly this set
250

Data Records in Yu et al. 2025, Scientific Data 12, 1080

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
Free registration

Create an account and accept terms online

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

Yu QS, Shan JY, Ma J, Gao G, Tao BZ, Qiao GY, Zhang JN, Wang T, Zhao YF, Qin XL, Yin YH. Multi-modal and Multi-view Cervical Spondylosis Imaging Dataset. Sci Data 12, 1080 (2025). https://doi.org/10.1038/s41597-025-05403-z

@article{yu2025mmcsd,
  title = {Multi-modal and Multi-view Cervical Spondylosis Imaging Dataset},
  author = {Yu, Qi-Shuai and Shan, Jing-Yang and Ma, Jie and Gao, Gan and Tao, Ben-Zhang and Qiao, Guang-Yu and Zhang, Jian-Ning and Wang, Ting and Zhao, Yong-Fei and Qin, Xiao-Lin and Yin, Yi-Heng},
  journal = {Scientific Data},
  volume = {12},
  pages = {1080},
  year = {2025},
  doi = {10.1038/s41597-025-05403-z}
}

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.