Automated CT bone segmentation using statistical shape modelling and local template matching.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Anatomy

Background:

• Accurate segmentation of bone structures from CT scans is critical for developing patient-specific implants using additive manufacturing.
• Current segmentation methods can struggle with complex or pathological bone regions, limiting their clinical applicability.

Purpose of the Study:

• To present an automated method for accurate CT bone segmentation that also provides anatomical correspondences.
• To validate the method's performance on challenging anatomical sites, including normal and osteoarthritic scapulae, healthy and atrophied mandibles, and orbital bones.

Main Methods:

• Development of an automated segmentation algorithm for CT bone data.
• Integration of anatomical correspondence propagation within the segmentation workflow.
• Evaluation using quantitative metrics such as surface distance and Dice coefficient on diverse bone datasets.

Main Results:

• The automated method achieved high accuracy in segmenting challenging bone regions.
• Average surface distances were approximately 0.5 mm, with Dice coefficients exceeding 90% across evaluated datasets.
• The method successfully provided anatomical correspondences during segmentation.

Conclusions:

• The developed automated segmentation method is highly accurate and reliable for complex bone structures.
• This approach facilitates direct anatomical measurements and design parameter extraction for personalized surgical instruments and patient-specific implants.
• The technique holds significant potential for advancing chair-side manufacturing of medical implants.