From predictive accuracy to clinical utility: Model interpretability and the missing bone mineral density data.

Abstract

This letter to the editor commends the study by Liu et al. on their machine learning model for predicting rib fractures but highlights two crucial challenges for its clinical adoption: the lack of model interpretability, which hinders clinical trust, and the omission of bone mineral density data, a key biological determinant of fracture risk. We suggest incorporating explainable AI techniques like SHAP/LIME and leveraging quantitative CT-based bone density measurements to enhance transparency, accuracy, and ultimately, the model's translational potential.