Overview

Schema Version

https://atlas.rsna.org/schemas/2025-11/dataset.json

Name

SNUH/BRMH hip AP radiograph cohorts

Link

https://pubmed.ncbi.nlm.nih.gov/35923378/

Indexing

Keywords: hip radiograph, osteoporosis, osteopenia, bone mineral density, DXA, deep learning, radiomics, Siamese network

Content: MK

RadLex: RID2641, RID35976, RID5389, RID12729, RID16657, RID10363, RID5388, RID12281

SNOMED: 64859006, 312894000

Author(s)

Sangwook Kim

Bo Ram Kim

Hee-Dong Chae

Jimin Lee

Sung-Joon Ye

Dong Hyun Kim

Sung Hwan Hong

Ja-Young Choi

Hye Jin Yoo

Organization(s)

Seoul National University Hospital (SNUH)

Seoul Metropolitan Government–Seoul National University Boramae Medical Center (BRMH)

Seoul National University Bundang Hospital

Ulsan National Institute of Science and Technology

Seoul National University College of Medicine

Institute of Radiation Medicine, Seoul National University Medical Research Center

Contact

Corresponding author: Hee-Dong Chae, MD; email: hdchae20@gmail.com; Department of Radiology, Seoul National University Hospital, 101 Daehak-ro, Jongno-gu, Seoul 03080, Republic of Korea.

Funding

Korea Health Technology R&D Project through the Korea Health Industry Development Institute, funded by the Ministry of Health & Welfare, Republic of Korea (grant HI20C2092); SNUH Research Fund (grant 04-2020-2060).

Ethical review

Institutional review boards of SNUH and BRMH approved this retrospective study; informed consent was waived.

Comments

Retrospective, IRB-approved study assembling hip/pelvis AP digital radiographs with paired DXA within 1 year for osteoporosis labeling; development at Seoul National University Hospital (SNUH) with temporal split; external testing at Seoul Metropolitan Government–Seoul National University Boramae Medical Center (BRMH).

Date

Published: 2022-05-25

References

[1] Kim S, Kim BR, Chae H-D, Lee J, Ye S-J, Kim DH, Hong SH, Choi J-Y, Yoo HJ. "Deep Radiomics–based Approach to the Diagnosis of Osteoporosis Using Hip Radiographs". Radiology: Artificial Intelligence. 2022. doi:10.1148/ryai.210212. PMID: 35923378. PMCID: PMC9344212.

Dataset

Motivation

To develop and validate deep radiomics models to diagnose osteoporosis using hip radiographs and assess added value to observer performance.

Sampling

Consecutive adult patients ≥18 years with hip/pelvis AP radiographs and DXA within 1 year; exclusions: fracture, hip surgery, osteonecrosis, bone tumors, suboptimal image quality.

Partitioning scheme

Temporal split at SNUH for training/validation vs internal test; geographic split for external test at BRMH.

Missing information

Exact file formats, per-partition series/image counts for internal test, and demographic breakdowns beyond those reported are not provided.

Relationships between instances

Multiple radiographs per patient in development cohort; both proximal femurs used simultaneously in a Siamese network; labels derived from lowest femoral neck/total femur DXA T score.

Noise

Variability in patient positioning and scanning parameters; addressed with preprocessing and model design; texture features showed lower generalizability.

External data

Geographically external test set from BRMH.

Confidentiality

Retrospective, IRB-approved; consent waived.

Sensitive data

Medical imaging data of patients; no explicit statement about PHI in image pixels.