The outcome of an automated assessment of trabecular pattern in intraoral radiographs as a fracture risk predictor

Abstract

Objectives: This study aims to investigate if automated analyses of the trabecular pattern in intraoral radiographs independently contribute to fracture risk assessment when other risk factors incorporated in the Fracture Risk Assessment Tool (FRAX) are taken into account. A secondary aim is to explore the correlation between the automated trabecular pattern assessment in intraoral radiographs and Trabecular Bone Score (TBS). Methods: A total of 567 intraoral radiographs from older females participating in a large population-based study (SUPERB) based in Gothenburg, Sweden, were selected to analyse trabecular pattern using semi-automated and fully automated software. Associations between trabecular pattern analysis and incident fractures were studied using Cox proportional hazard model, unadjusted and adjusted for FRAX risk factors (previous fracture, family history of hip fracture, smoking, corticosteroids, rheumatoid arthritis, without and with bone mineral density (BMD) of the femoral neck). In addition, the correlation between trabecular pattern analysis and TBS of the lumbar spine was investigated using Pearson correlation analysis. Results: Neither the unadjusted nor the adjusted trabecular pattern analysis in intraoral radiographs was significantly associated with any fracture or major osteoporotic fracture (MOF). A weak correlation was found between semi-automated trabecular pattern analysis and TBS. No correlation was found between the fully automated trabecular pattern analysis and TBS. Conclusions: The present study shows that semi-automated and fully automated digital analyses of the trabecular pattern in intraoral radiographs do not contribute to fracture risk prediction. Furthermore, the study shows a weak correlation between semi-automated trabecular pattern analysis and TBS.