Explainable machine learning for osteoporosis detection in patients with osteopenia: model development and validation...

Area of Science:

• Bone health research
• Medical diagnostics
• Machine learning in healthcare

Background:

• Osteopenia is a precursor to osteoporosis, but differentiating them in low bone mass individuals is challenging.
• Dual-energy X-ray absorptiometry (DXA) is definitive but not always accessible.
• Need for alternative screening methods using readily available clinical data.

Purpose of the Study:

• To develop and validate a machine learning algorithm for discriminating osteoporosis from osteopenia.
• To utilize easily accessible clinical data for this diagnostic model.
• To provide a practical screening tool for primary care settings.

Main Methods:

• Retrospective diagnostic study of 1,203 Asian adults with low bone mass.
• Trained and validated 11 machine learning algorithms to classify osteoporosis (T-score ≤ -2.5) vs. osteopenia (T-score -2.5 to -1.0).
• Evaluated model performance using Area Under the Curve (AUC), SHAP analysis, nomogram calibration, and Decision Curve Analysis (DCA).

Main Results:

• Linear Discriminant Analysis model showed superior performance with a cross-validated AUC of 0.738 and independent validation AUC of 0.710.
• Key predictors included waist-to-height ratio, body weight, serum uric acid, age, and alkaline phosphatase.
• The model demonstrated excellent calibration and positive net benefit across various risk thresholds.

Conclusions:

• Developed a practical, interpretable tool for discriminating osteoporosis from osteopenia using routine clinical data.
• This algorithm can serve as a preliminary screening tool in primary care.
• Aids in identifying individuals with low bone mass who require further definitive bone density testing.