Digital Imaging and Communications in Medicine (DICOM) metadata lack standardization and are incomplete. Therefore, there have has been inefficiencies in the data retrieval process with DICOM image data. The structure and terminology system of R-CDM was designed to standardize DICOM metadata. Furthermore, a deep learning image classifier was developed to improve the quality of the metadata in the DICOM file because incorrect metadata is a common finding. As a proof of concept, after converting medical image data from Ajou University Hospital into a standardized form of R-CDM, a deep learning image classifier was used to improve the quality of the standardized metadata. Finally, by connecting R-CDM with Observational Medical Outcomes Partnership (OMOP)-CDM, researchers can easily search for the desired type of medical image data for a specific patient cohort. Furthermore, it is possible to efficiently secure large-scale medical image data from multiple institutions for developing deep learning models.