Predicting Thoracolumbar Vertebral Osteoporotic Fractures: Value Assessment of Chest CT-Based Machine Learning

Comprehensive Summary

This study evaluates the use of routine chest CT scans in conjunction with a machine learning model to predict osteoporotic vertebral fractures (OVFs) in the thoracolumbar vertebral bodies. Researchers analyzed chest CT imaging data, along with demographic and laboratory data, to identify features associated with fracture risk. A total of 162 patients were included in the study, with 113 assigned to the training cohort and 49 assigned to the testing cohort. Volumetric bone mineral density (vBMD) and hemoglobin (HGB) were identified as independent risk factors in the clinical model, with odds ratios of 0.95 and 0.96, respectively. Three predictive models were developed and tested, including a clinical model, a radiomics-based model, and a combined nomogram model. The Nomogram model, which incorporated Radscore, vBMD, and HGB, displayed strong discriminatory power in predicting the risk of thoracolumbar vertebral fragility fractures.

Outcomes and Implications

Currently, OVFs are associated with substantial treatment costs, reduced quality of life, and increased mortality, underscoring the need for early identification and prevention. The ability of machine learning models to integrate multidimensional clinical and imaging data represents a valuable advancement in fracture risk prediction. However, the findings of this study are limited by a relatively small sample size and the need for additional validation.