Accuracy of femur reconstruction from sparse geometric data using a statistical shape model

Area of Science:

• Biomechanical Engineering
• Medical Imaging
• Computational Anatomy

Background:

• Reconstructing complete femur geometry from limited medical images is crucial for accurate biomechanical models of the hip and knee.
• Existing methods often rely on statistical shape models to estimate full femur geometry from sparse data.
• The impact of varying data sparsity on reconstruction accuracy remains under-investigated.

Purpose of the Study:

• To systematically compare statistical shape model and linear scaling techniques for femur reconstruction.
• To evaluate the influence of different proportions of proximal and distal femur data on reconstruction accuracy.
• To assess reconstruction error using both surface-to-surface distance and anatomical coordinate system deviations.

Main Methods:

• Utilized statistical shape models and linear scaling for femur surface reconstruction.
• Employed varying amounts of proximal and distal femur geometry, alongside morphometric and landmark data.
• Quantified errors via surface-to-surface measurements and anatomical axis deviations.

Main Results:

• Shape model reconstruction yielded a mean surface error of 1.8 mm and anatomical axis error <0.15° using partial proximal femur data.
• Linear scaling resulted in significantly higher errors (19.1 mm surface, 2.7-5.6° axis error).
• Varying the amount of partial proximal or distal data showed negligible impact on shape model accuracy.

Conclusions:

• Statistical shape models offer superior accuracy for full femur reconstruction compared to linear scaling.
• Appropriate sparse geometric data enables highly accurate femur geometry estimation for biomechanical applications.
• Shape models provide a robust solution for establishing accurate boundary conditions in hip and knee biomechanical models.