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

Area of Science:

• Medical imaging and artificial intelligence
• Orthopedic research
• Radiology

Background:

• Machine learning models show promise for predicting rib fractures.
• Clinical adoption of these models is hindered by a lack of transparency and incomplete data.
• Bone mineral density is a critical factor in fracture risk.

Purpose of the Study:

• To evaluate the clinical applicability of a machine learning model for rib fracture prediction.
• To identify key areas for improvement in the model's design and data inputs.
• To propose methods for enhancing the model's trustworthiness and accuracy.

Main Methods:

• Review and critique of existing machine learning models for rib fracture prediction.
• Discussion of the importance of model interpretability in clinical decision-making.
• Recommendation for integrating quantitative CT-based bone density measurements.

Main Results:

• The reviewed machine learning model for rib fracture prediction has limitations.
• Lack of interpretability reduces clinical trust and adoption.
• Exclusion of bone mineral density data limits predictive accuracy.

Conclusions:

• Explainable AI (Artificial Intelligence) techniques such as SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) should be incorporated.
• Quantitative CT-based bone density measurements are essential for improving fracture prediction.
• These enhancements will increase the model's transparency, accuracy, and translational potential in clinical settings.