Application of machine learning classifiers for microcomputed tomography data assessment of mouse bone...

Area of Science:

• Biomedical Engineering
• Orthopedics
• Materials Science

Background:

• Current analysis of bone microarchitecture from micro-computed tomography (microCT) relies on traditional statistical tests (e.g., ANOVA, t-tests, Mann-Whitney U).
• These univariate methods are susceptible to Type 1 errors due to multiple comparisons, potentially leading to biased significance reporting.

Purpose of the Study:

• To introduce and evaluate machine learning classification methods as a complementary approach to traditional statistical analyses for microCT data.
• To demonstrate how machine learning can minimize bias and Type 1 errors in assessing bone structure.

Main Methods:

• Application of unsupervised k-means cluster analysis for bone microarchitecture assessment.
• Utilization of supervised Support Vector Machine (SVM) classification for analyzing cortical bone structure.
• Simultaneous evaluation of multiple microCT measures within the machine learning frameworks.

Main Results:

• Machine learning methods (k-means, SVM) simultaneously process numerous microCT measures, reducing bias.
• These techniques effectively minimize Type 1 error rates associated with multiple comparisons in statistical testing.
• Machine learning provides a more robust assessment of microCT-derived bone measures compared to univariate tests.

Conclusions:

• Machine learning classification offers a powerful and complementary approach to traditional statistical tests for microCT analysis.
• K-means clustering and SVM classification enhance the reliability and robustness of bone microarchitecture and cortical bone structure assessments.
• Adopting machine learning can lead to more accurate and less biased interpretations of microCT data in bone research.