Development and validation of a management system and dataset quality assessment tool for the Radiology Common Data Model (R_CDM): A case study in liver disease

Tae-Hoon Kim; SiHyeong Noh; Youe Ree Kim; ChungSub Lee; Ji Eon Kim; Chang-Won Jeong; Kwon-Ha Yoon

doi:10.1016/j.ijmedinf.2022.104759

Development and validation of a management system and dataset quality assessment tool for the Radiology Common Data Model (R_CDM): A case study in liver disease

Int J Med Inform. 2022 Apr 1:162:104759. doi: 10.1016/j.ijmedinf.2022.104759. Online ahead of print.

Authors

Tae-Hoon Kim¹, SiHyeong Noh¹, Youe Ree Kim², ChungSub Lee¹, Ji Eon Kim¹, Chang-Won Jeong³, Kwon-Ha Yoon⁴

Affiliations

¹ Medical Convergence Research Center, Wonkwang University, Iksan 54538, Republic of Korea.
² Department of Radiology, Wonkwang University School of Medicine and Wonkwang University Hospital, Iksan 54538, Republic of Korea.
³ Medical Convergence Research Center, Wonkwang University, Iksan 54538, Republic of Korea. Electronic address: mediblue@wku.ac.kr.
⁴ Medical Convergence Research Center, Wonkwang University, Iksan 54538, Republic of Korea; Department of Radiology, Wonkwang University School of Medicine and Wonkwang University Hospital, Iksan 54538, Republic of Korea. Electronic address: khy1646@wku.ac.kr.

PMID: 35390589
DOI: 10.1016/j.ijmedinf.2022.104759

Abstract

Background: The Observational Medical Outcomes Partnership-Common Data Model (OMOP-CDM), a distributed research network, has low clinical data coverage. Radiological data are valuable, but imaging metadata are often incomplete, and a standardized recording format in the OMOP-CDM is lacking. We developed a web-based management system and data quality assessment (RQA) tool for a radiology_CDM (R_CDM) and evaluated the feasibility of clinically applying this dataset.

Methods: We designed an R_CDM with Radiology_Occurrence and Radiology_Image tables. This was seamlessly linked to the OMOP-CDM clinical data. We adopted the standardized terminology using the RadLex playbook and mapped 5,753 radiology protocol terms to the OMOP vocabulary. An extract, transform, and load (ETL) process was developed to extract detailed information that was difficult to extract from metadata and to compensate for missing values. Image-based quantification was performed to measure liver surface nodularity (LSN), using customized Wonkwang abdomen and liver total solution (WALTS) software.

Results: On a PACS, 368,333,676 DICOM files (1,001,797 cases) were converted to R_CDM chronic liver disease (CLD) data (316,596 MR images, 228 cases; 926,753 CT images, 782 cases) and uploaded to a web-based management system. Acquisition date and resolution were extracted accurately, but other information, such as "contrast administration status" and "photography direction", could not be extracted from the metadata. Using WALTS, 9,609 pre-contrast axial-plane abdominal MR images (197 CLD cases) were assigned LSN scores by METAVIR fibrosis grades, which differed significantly by ANOVA (p < 0.001). The mean RQA score (83.5) indicated good quality.

Conclusion: This study developed a web-based system for management of the R_CDM dataset, RQA tool, and constructed a CLD R_CDM dataset, with good quality for clinical application. Our management system and R_CDM CLD dataset would be useful for multicentric and image-based quantification researches.

Keywords: Chronic liver disease (CLD); Metadata; Radiology_common data model (R_CDM); Standardization.