Main Session
Sep 30
PQA 09 - Hematologic Malignancies, Health Services Research, Digital Health Innovation and Informatics

3623 - Development of the Ontology for Osteoradionecrosis of the Jaws (ORNJ-O) for Classification System Harmonization

04:00pm - 05:00pm PT
Hall F
Screen: 22
POSTER

Presenter(s)

Anthony Giovi, MS, BS Headshot
Anthony Giovi, MS, BS - The University of Texas MD Anderson Cancer Center, Houston, TX

A. Giovi1, A. M. Masci2, L. Humbert-Vidan1, S. Y. Lai3, K. A. Hutcheson3, M. S. Chambers3, E. Watson4, L. V. van Dijk5, C. D. Fuller1, and A. C. Moreno6; 1Department of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX, 2The University of Texas MD Anderson Cancer Center, Houston, TX, 3Department of Head and Neck Surgery, The University of Texas MD Anderson Cancer Center, Houston, TX, 4Department of Dental Oncology and Maxillofacial Prosthetics, Princess Margaret Cancer Centre, Toronto, ON, Canada, 5Department of Radiation Oncology, University Medical Center Groningen, Groningen, Netherlands, 6Department of Radiation Oncology, University of Texas MD Anderson Cancer Center, Houston, TX

Purpose/Objective(s): Osteoradionecrosis of the jaw (ORNJ) is an iatrogenic condition that results from changes in vascularity of irradiated bone leading to bone necrosis or death. To date, there are at least 16 classification systems for ORNJ, leading to inconsistencies in reporting and likely underdiagnosis of ORNJ and its severity. The objective of this work was to integrate and harmonize the conceptual knowledge of ORNJ across various classification systems, ensuring a unified and comprehensive ontological framework. This integration is essential to establish a standardized and cohesive understanding of ORNJ classification.

Materials/Methods: A literature review was performed to identify 16 classification systems for ORNJ published from 1983 to 2024. Staging/grading criteria for each system was manually reviewed, extracted, and approved by 2 reviewers. Semantic triples, representing subject-predicate-object relations, were formulated and converted to explicit classes and properties using the ontology editor software, Protégé. UBERON, an anatomical ontology, and SNOMED CT, a standardized clinical terminology, were referenced as primary standards. The final ORNJ ontology (ORNJ-O) was serialized in RDF/XML format.

Results: The ORNJ ontology consists of a total of 133 classes, 2 data properties, 10 object properties, and 175 logical axioms. Common classes identified include BoneExposure, BoneNecrosis, PathologicalFracture which were represented in 56% (9/16), 56% (9/16), and 50% (8/16) of reviewed systems, respectively. Other classes defined include Symptoms, Therapy, AnatomicalCharacteristics, and ClassificationSystem. A quarter of classification systems considered time (in months) as a critical concept but varied in a specific threshold range. Relationships were explicitly defined through properties including hasClinicalFindings, hasRadiographicFindings, and hasLengthInCM. Using ORNJ-O, an individual with bone exposure and pain lasting more than 3 months with radiographic evidence of alveolar bone necrosis was semantically inferred to belong to different stages across multiple systems: to ClinRad: Stage 1, Epstein: Stage 2a, Morton Simpson: Minor stage, and Glanzman: Grade 1.

Conclusion: By explicitly defining clinical and radiographic concepts and relationships associated with ORNJ and existing classification systems, ORNJ-O was engineered to optimize data harmonization and interoperability. Not all stages could be harmonized, reflecting the current challenges in comparing system-varied ORNJ cohorts over time. However, ORNJ-O comprehensively defines all features of the recently endorsed ClinRad system, thereby enabling standardized, prospective data collection.