Leveraging Mcode to Enable Interoperable Ontology for Oncology Digital Twins
Abstract
Purpose
Clinical centres currently employ institution-specific methods for patient imaging (including imaging protocols and preprocessing methods), register, and store information. This lack of semantic standardization creates significant barriers to cross-institutional learning and interoperability, introducing data bias due to a preponderance of single-centre studies. While data standards exists, their adoption remains limited. To solve this issue, we present a semantic data model—an ontology—to formally define concepts and standardize oncology expertise, enabling bias-aware digital twin development.
Methods
We developed our ontology leveraging the Minimal Common Oncology Data Elements (mCODE) standard, arose from an international initiative focused on six core thematic groups: Patient, Disease, Genomics, Treatment, Outcomes, and Assessment. We followed METHONTOLOGY, a standardized ontology engineering methodology encompassing five steps: domain and scope definition, reuse ontologies, conceptual model developing, ontology implementation, and evaluation. To capture complexity of multimodal patient data, we extended the mCODE model with three new groups: Patient Medical Images, Radiomics, and Dosiomics. The ontology reuses and aligns with established biomedical standards, including SNOMED CT, RadLex, O3, and O-RAW, and was implemented using Protégé and the Web Ontology Language (OWL).
Results
We defined the ontology domain scope to encompass radiation oncology and medical physics domain. We fully developed conceptual maps for each mCODE thematic group and extensions, building on the initial study by El Ghosh et al. (2024). The proposed ontology formally integrates clinical, genomic, imaging, radiomics, and dosiomics data within a unified semantic framework. The ontology structure has been implemented in OWL, with class hierarchies and relationships defined for core mCODE elements and extensions; linking clinical context to images and quantitative features. Our next step consists of evaluating our ontology using glioma real-world data.
Conclusion
Our mCODE-based ontology aligns biomedical standards and integrates multimodal oncology data; It enables semantic interoperability and provides a foundational infrastructure for scalable oncology digital twins.