Scoliosis1K: A Large-Scale Dataset for Video-Based Scoliosis Screening

¹Southern University of Science and Technology, ²Shenzhen University

Highlights in Scoliosis1K

Large-Scale & Real-World: 1,493 walking video sequences (447,000 frames) from 1,050 adolescents collected in a real-world school-based screening program.
Expert-Annotated Labels: Each subject was evaluated on-site by medical professionals and categorized as positive (scoliosis), neutral (borderline), or negative (non-scoliosis) based on standard screening protocols.
Privacy-Preserving Representations: The dataset currently includes silhouette and 2D pose representations derived from RGB videos, ensuring participant privacy. Additional representations will be released in the future.

How to get Scoliosis1K

Please follow the steps below to access and use the dataset:

Step 1: Download

Choose a download link:

OneDrive
BaiduYun (Password: 54iu)

Step 2: Request Access

Download the Dataset Agreement.
Sign it and email to 12331257@mail.sustech.edu.cn with subject: [Scoliosis1K Dataset Application].
You will receive the decompression password upon approval.
Then, follow the usage guide to extract and use the dataset.

Important: This dataset is released for ACADEMIC USE ONLY. Please ensure compliance with the agreement terms and conditions.

Related Publications

All publications using "Scoliosis1K" dataset should cite the following papers:

Zirui Zhou^†, Junhao Liang^†, Zizhao Peng, Chao Fan, Fengwei An, Shiqi Yu^*, "Gait Patterns as Biomarkers: A Video-Based Approach for Classifying Scoliosis", International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI), 2024. [PDF] [Code]
Zirui Zhou, Zizhao Peng, Dongyang Jin, Chao Fan, Fengwei An, Shiqi Yu^*, "Pose as Clinical Prior: Learning Dual Representations for Scoliosis Screening", International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI), 2025. [PDF] [Code]

^* Corresponding author, ^† Equal contribution (co-first authors)

BibTeX

@inproceedings{zhou2024gait,
      title={Gait Patterns as Biomarkers: A Video-Based Approach for Classifying Scoliosis},
      author={Zhou, Zirui and Liang, Junhao and Peng, Zizhao and Fan, Chao and An, Fengwei and Yu, Shiqi},
      booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
      pages={284--294},
      year={2024},
      organization={Springer}
    }
@inproceedings{zhou2025gait,
  title={Pose as Clinical Prior: Learning Dual Representations for Scoliosis Screening},
  author={Zhou, Zirui and Peng, Zizhao and Jin, Dongyang and Fan, Chao and An, Fengwei and Yu, Shiqi},
  booktitle={International Conference on Medical Image Computing and Computer-Assisted Intervention},
  year={2025},
  organization={Springer}
}