Artificial intelligence for medication-related osteonecrosis of the jaw: a scoping review.

OBJECTIVES

To describe the current state of research on the application of artificial intelligence (AI) to the prediction, diagnosis, and management of medication-related osteonecrosis of the jaw (MRONJ).

STUDY

DESIGN

A scoping review was conducted using PubMed/MEDLINE, Scopus, and Web of Science databases through March 1, 2024, and updated to December 31, 2024, according to PRISMA-SCR guidelines. Data on AI tasks, models, patient information, datasets, and performance metrics were extracted from eligible studies.

RESULTS

Eight studies met the inclusion criteria, focusing on MRONJ onset prediction (n = 5), diagnosis (n = 2), and patient education (n = 1). For onset prediction, machine learning models (support vector machines, random forests, gradient boosting machines) achieved area under the curve values of 0.793-0.973. For diagnosis, deep learning models using radiographic images achieved 96% accuracy with 93% precision and recall. For patient education, a large language model was evaluated with moderate to high response quality.

CONCLUSIONS

While AI shows promise in the prediction and diagnosis of MRONJ, challenges related to data quality, validation, and clinical integration need to be addressed. Future research should focus on developing standardized, explainable AI models and establishing implementation guidelines for clinical practice.