Demystifying machine learning approaches in digital bone imaging using microCT and HRpQCT.

Area of Science:

• Biomedical Imaging
• Computational Biology
• Orthopedics

Background:

• Micro-computed tomography (microCT) and high-resolution peripheral quantitative computed tomography (HRpQCT) provide detailed 3D bone imaging.
• Radiomics extracts quantitative bone microarchitecture data, but conventional analysis struggles with complex, nonlinear relationships and data integration.
• Integrating imaging data with biomechanics, histology, and clinical data presents significant analytical challenges in bone research.

Purpose of the Study:

• To review machine learning (ML) applications in preclinical and clinical bone research using microCT and HRpQCT.
• To synthesize the intersection of bone research and ML via scientometric and bibliometric analyses using SciNetX software.
• To provide a foundational understanding of ML model development and interpretation for bone imaging research.

Main Methods:

• Literature review of ML applications in bone imaging (microCT, HRpQCT).
• Scientometric and bibliometric analysis using SciNetX for visualizing research trends.
• Explanation of ML model development and interpretation principles relevant to bone research.

Main Results:

• ML offers powerful tools to enhance bone spatial resolution and accelerate image segmentation.
• ML can uncover hidden patterns and relationships within high-dimensional bone imaging data.
• Insights into ML model inputs can identify potential biological phenotypes or therapeutic targets.

Conclusions:

• Machine learning presents significant opportunities to advance bone research by overcoming limitations of conventional imaging analysis.
• Integrating ML with microCT and HRpQCT data can lead to deeper understanding of bone structure and quality.
• This review serves as a guide for researchers incorporating ML into digital imaging techniques for bone studies.