A NURBS-based technique for subject-specific construction of knee bone geometry

Area of Science:

• Biomedical Engineering
• Medical Imaging
• Computational Anatomy

Background:

• Accurate subject-specific finite element (FE) models of knee joint bones are crucial for biomechanical analysis.
• Existing methods for constructing bone geometries from CT images often struggle with precise endosteal surface and thin cortical bone representation.

Purpose of the Study:

• To introduce a semi-automatic non-uniform rational B-spline (NURBS) technique for rapid and accurate construction of knee bone geometries from CT images.
• To improve the capture of endosteal surfaces and cortical bone thickness, particularly in complex regions like the metaphysis and epiphysis.

Main Methods:

• A semi-automatic NURBS-based technique combining edge extraction and CAD surface generation was developed.
• CT data from transverse and sagittal planes were integrated to address surface generation challenges at the femoral condyles.
• Cortical thickness was estimated by analyzing surrounding bone structure to accommodate thin cortical bone.

Main Results:

• The proposed NURBS technique accurately captured periosteal surfaces, comparable to voxel mesh methods.
• Unlike voxel mesh models, the NURBS technique accurately depicted the metaphyseal and epiphyseal endosteum, overcoming CT contrast limitations.
• The NURBS method achieved cortical bone thickness accuracy within 2 pixel lengths, significantly outperforming voxel methods in complex bone regions.

Conclusions:

• The semi-automatic NURBS technique offers a significant advancement in constructing accurate subject-specific knee bone geometries from CT data.
• This method provides superior endosteal surface and cortical bone thickness definition, essential for reliable FE modeling of the knee joint.
• The technique's ability to handle thin cortical bone and complex anatomical regions enhances its utility in orthopedic research and clinical applications.