Kyphosis Dataset

The first open-access dataset of biplanar EOS images centered on kyphosis. This collection features radiologist-annotated spinopelvic parameters designed to benchmark automated spinal deformity assessment algorithms.

Data Description

Data was collected from patients evaluated between 2018 and 2022 at Shafa Yahyaian Hospital, Tehran, Iran. The dataset addresses the gap in publicly available low-dose, full-body spinal imaging data needed for musculoskeletal AI research.

Imaging Data

● 100 Biplanar Scans: Anonymized full-body EOS images.
● Formats: DICOM format with raw pixel data.

Expert Annotations

● Cobb Angles (Kyphosis/Lordosis)
● Pelvic Parameters (Incidence, Tilt, SVA)

Request Data Access

This dataset is available strictly for non-commercial academic research purposes. Individual researchers must fill out the form below and sign the Data Use Agreement (DUA).

📄Download Data Use Agreement (.docx)

Researcher Details

Name

Institution of Affiliation

Position

City

Country

Address

Contact Number

Aim of Access Please specify your research goal.

Upload Signed Agreement

Please upload the signed file (.docx/.pdf).

By submitting this form, I acknowledge my understanding of and agreement to the terms and conditions associated with accessing and using this dataset for non-commercial academic research purposes.

Lateral View
AP View

Sagittal plane visualization

Coronal plane visualization

Open Access Sample

Download a small, anonymized sample batch (5 patients) to verify file formats and structure. No DUA required for sample data.

Download Sample Data (.zip)