Application of machine learning on health examination data for predicting the decrease of bone mineral density

Area of Science:

• Computational medicine
• Biomedical data science
• Osteoporosis research

Background:

• Diminished bone density poses significant health risks and economic burdens.
• Early identification and preventative strategies are crucial for patient well-being.
• Developing precise forecasting tools for bone mineral density loss is a key healthcare objective.

Purpose of the Study:

• To develop and evaluate machine learning algorithms for predicting bone mineral density loss.
• To compare the efficacy of five distinct machine learning models in forecasting osteoporosis risk.
• To identify the most accurate model for clinical application in bone health assessment.

Main Methods:

• Utilized health examination data from 11,132 individuals aged 40+ (2022-2024).
• Applied five machine learning algorithms: k-nearest neighbor (KNN), random forest (RF), support vector machine (SVM), artificial neural network (ANN), and logistic regression (LR).
• Assessed model performance using accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUROC).

Main Results:

• The random forest (RF) model achieved the highest AUROC scores (0.918 for males, 0.922 for females).
• RF also demonstrated the highest accuracy (0.88 for males, 0.85 for females).
• Artificial neural network (ANN) and support vector machine (SVM) also showed strong predictive capabilities.

Conclusions:

• Machine learning, particularly the RF model, offers a precise method for predicting bone density reduction.
• These predictive models can enhance the prevention, identification, and early intervention of osteoporosis.
• Clinical implementation of these algorithms can significantly improve patient outcomes and reduce healthcare costs.