Overview

Schema Version

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

Name

Samsung Medical Center Chest Radiograph dataset

Link

https://pubmed.ncbi.nlm.nih.gov/38446041/

Indexing

Keywords: Artificial Intelligence, Conventional Radiography, Humerus, Machine Learning, Shoulder, Tumor, Humeral metastasis, Chest radiograph

Content: CH, MK, OI

RadLex: RID41946, RID41945, RID35179, RID39282

SNOMED: 767813008, 94222008, 115239009

Author(s)

Harim Kim

Kyungsu Kim

Seong Je Oh

Sungjoo Lee

Jung Han Woo

Jong Hee Kim

Yoon Ki Cha

Kyunga Kim

Myung Jin Chung

Organization(s)

Department of Radiology, Samsung Medical Center, Sungkyunkwan University School of Medicine

Medical AI Research Center, Samsung Medical Center

Department of Data Convergence and Future Medicine, Sungkyunkwan University School of Medicine

Samsung Advanced Institute for Health Sciences & Technology (SAIHST), Sungkyunkwan University

License

URL: https://pubs.rsna.org/doi/10.1148/ryai.230094

Contact

Corresponding author: Myung Jin Chung (email: ude.ukks@gnuhcjm)

Funding

Korea Medical Device Development Fund (202011B08-02, KMDF_PR_20200901_0014-2021-02); Technology Innovation Program (20014111) funded by MOTIE (Korea); Future Medicine 20*30 Project of Samsung Medical Center (SMX1210791).

Ethical review

Institutional Review Board approval obtained; informed consent waived (IRB no. 2022–08–099).

Comments

Retrospective single-center dataset of chest radiographs (PA/AP; fixed or portable) with CT-proven labels (normal vs humeral tumor) collected January 2000–December 2021.

Date

Updated: 2024-02-15

Published: 2024-03-06

Created: 2000-01-01

References

[1] Kim H; Kim K; Oh SJ; Lee S; Woo JH; Kim JH; Cha YK; Kim K; Chung MJ. "AI-assisted Analysis to Facilitate Detection of Humeral Lesions on Chest Radiographs". Radiology: Artificial Intelligence. 2024-03-06. doi:10.1148/ryai.230094. PMID: 38446041. PMCID: PMC11140509.

Dataset

Motivation

To develop and evaluate an AI system for detecting humeral tumors on chest radiographs and to assess impact on reader performance.

Sampling

CRs with both humeral heads included; inclusion January 2000–December 2021; CT-proven tumor or CT-proven normal at a single tertiary center in South Korea.

Partitioning scheme

Data split into: Training set; Holdout test set 1 (tumor and normal); Holdout test set 2 (normal only); Additional training set for humerus segmentation.

Missing information

Exact counts for holdout test set 1 not fully provided in the text; image resolution and de-identification specifics not described.

Relationships between instances

Multiple images per patient may be present; both left and right humeri evaluated per image; some patients contributed both normal and tumor images (normal at least 3 years apart from tumor diagnosis).

Noise

Suboptimal-quality images and those causing technical errors (e.g., DICOM load errors) were excluded; variability in humeral head position noted.

External data

CT reports and CT examinations used as the reference standard for labeling (normal vs humeral tumor).

Confidentiality

Retrospective clinical imaging data from a single tertiary medical center; IRB approved with waiver of consent.

Re-identification

Not described; retrospective single-center CRs. De-identification specifics not stated.

Sensitive data

Medical images (DICOM) with potential PHI; specifics of PHI removal not stated.