Open-source pipeline for automatic segmentation and microstructural analysis of murine knee subchondral bone

Area of Science:

• Biomedical Engineering
• Orthopedics
• Radiology

Background:

• Micro-computed tomography (μCT) is crucial for assessing bone density and microstructure in preclinical models.
• Manual segmentation of subchondral bone regions for volumetric analysis is time-consuming and limits throughput.
• Automated analysis platforms exist but often require manual segmentation as a prerequisite.

Purpose of the Study:

• To develop an automated end-to-end pipeline for segmenting and analyzing subchondral bone microstructure in mouse proximal knee μCT images.
• To compare the accuracy and agreement of an automated segmentation U-Net architecture with manual segmentation methods.
• To evaluate microstructural parameter quantification using the ITK BoneMorphometry library and CTAn software.

Main Methods:

• A U-Net architecture was trained on μCT scans from 62 mouse knees (healthy and arthritic).
• Automated segmentations were used with original scans for microstructural analysis via ITK and CTAn pipelines.
• Key parameters including bone volume (BV), total volume (TV), BV/TV, trabecular number (TbN), trabecular thickness (TbTh), trabecular separation (TbSp), and bone surface density (BSBV) were compared.

Main Results:

• High correlation (R = 0.88–0.98 for ITK, R = 0.91–0.98 for CTAn) was observed between manual and U-Net automated segmentation pipelines.
• Good agreement was found for most parameters between ITK and CTAn, with minor discrepancies in TbN and TbSp due to differing methodologies.
• Automated segmentations resulted in slightly lower average values for BV, TV, and BV/TV, but these differences were not influenced by mean ROI values.

Conclusions:

• An integrated, automated pipeline for subchondral bone segmentation and microstructural analysis was successfully developed.
• The open-source pipeline facilitates efficient analysis of large μCT datasets, potentially standardizing trabecular bone microstructural assessment.
• This automated approach enhances throughput and reproducibility in preclinical bone research.