Metadados em Imagem Médica: Como promover o seu potencial na geração de Evidência em Investigação Clínica

Abstract

Introduction: Generalized Healthcare digitization brings major challenges in data management and sharing. Standards for modeling, encoding, communicating and storing are key in addressing these challenges, ensuring effective information management and usage. In medical imaging, the Digital Imaging and Communications in Medicine (DICOM) standard is widely used, enabling the storage of both images and metadata with great potential for research. Nonetheless, its metadata remains largely underutilized for clinical research purposes. Meanwhile, the Observational Medical Outcomes Partnership Common Data Model (OMOP-CDM) is gaining traction in observational research. This study aims to explore the potential of medical imaging metadata for clinical research, addressing the challenges and opportunities of its reuse, including its integration into the OMOP-CDM. Methods: The study followed a sequential proof-of-concept approach. The first phase consisted of a state-of-the-art analysis and the identification of relevant DICOM metadata in open-access datasets. This was followed by the design of an architecture based on selected open-source tools (Orthanc, Apache NiFi, PostgreSQL), implemented in a reproducible environment using devcontainers. The conceptual mapping of metadata to the OMOP-CDM was performed with the support of OHDSI tools (White Rabbit and Rabbit in a Hat). Finally, validation of the solution included technical verification of data integrity and a demonstration of its utility through descriptive analysis. Results: A modular and reproducible architecture, orchestrated via Docker, was implemented, enabling the development of an ETL pipeline. As a result of this process, metadata from 77,212 DICOM instances corresponding to 117 unique studies were analysed, identifying a total of 807 distinct tags. Descriptive analysis revealed a core of 33 tags present in 100% of studies. In order to address the aim of study although not prevalent in all studies a set of 38 tags particularly related to the domain were seletect for further analisis.Following the extraction and transformation of metadata, White Rabbit and Rabbit in a Hat were used to examine the data structure and design the mapping to OMOP-CDM tables, culminating in the generation of a formal ETL specification. Conclusions: This work explores the integration of DICOM metadata into the OMOP-CDM as a contribution towards strengthening interoperability and promoting observational research. The main metadata elements suitable for reuse in clinical studies are identified, the challenges inherent to integration are discussed, and a conceptual and methodological framework for the harmonisation of medical imaging data is proposed. ● Keywords: Metadata, DICOM, OMOP; Medical Image.