A deep learning-based framework for standardized analysis of trabecular bone compartments from micro-CT imaging data...

Area of Science:

• Skeletal Biology
• Medical Imaging
• Computational Biology

Background:

• Preclinical skeletal research relies on micro-computed tomography (micro-CT) for bone analysis in murine models.
• Inconsistent definitions of volumes of interest (VOIs) in trabecular bone compartments hinder reproducibility and statistical interpretation.
• Standardized analysis is crucial for evaluating interventions in bone remodeling and disease progression.

Purpose of the Study:

• To develop and validate a deep learning framework for automated segmentation of trabecular bone compartments in mouse tibia micro-CT scans.
• To establish standardized volumes of interest (VOIs) for consistent analysis across different experimental conditions.
• To improve the reproducibility and reliability of preclinical skeletal research.

Main Methods:

• A deep learning framework was developed for automated analysis of micro-CT scans (5 µm voxel size) of the epiphyseal-metaphyseal region in mouse tibia.
• A 2D slice-wise classification model combined with regional probability distribution identified four anatomical compartments: epiphyseal bone, growth plate, primary spongiosa, and secondary spongiosa.
• A deep learning-based segmentation model was used to segment trabecular bone within these compartments, followed by morphological and statistical analysis.

Main Results:

• The classification model achieved high performance across multiple datasets (mean F1-scores ranging from 0.92 to 0.99).
• The method demonstrated strong generalizability on an external dataset, with mean F1-scores ranging from 0.92 to 1.0.
• Statistical equivalence was achieved within 0.05 mm for compartment landmark detection.
• The automated analysis facilitated consistent comparisons across diverse experimental conditions, including pharmacological treatments and mechanical loading.

Conclusions:

• The developed deep learning framework provides a consistent, robust, and automated tool for analyzing micro-CT scans of trabecular bone in the mouse tibia.
• This standardization of VOI extraction and segmentation enhances reproducibility in preclinical skeletal research.
• The method facilitates advancements in understanding bone remodeling and disease progression by enabling reliable comparisons within and between trabecular compartments.