An articulated shape model to predict paediatric lower limb bone geometry using sparse landmarks.

Area of Science:

• Biomechanics
• Pediatric Orthopedics
• Medical Imaging

Background:

• Current methods for pediatric musculoskeletal modeling include image-based models (accurate but time-consuming) and linear scaling (faster but less accurate).
• There is a need for methods that are both fast and accurate for creating pediatric musculoskeletal models.

Purpose of the Study:

• To develop and validate an articulated shape model for predicting pediatric lower limb bone geometry.
• To compare the accuracy of the articulated shape model against linear scaling methods.

Main Methods:

• Developed an articulated shape model using data from 333 children (aged 4-18 years).
• The model predicts bone geometry from eight common motion capture landmarks.
• Evaluated bone surface errors and clinical bone measurements against linear scaling.

Main Results:

• The articulated shape model achieved lower bone surface root mean squared errors (e.g., 1.72 ± 0.51 mm for tibia/fibula) compared to linear scaling (e.g., 4.39 ± 0.86 mm).
• Clinical bone measurements showed low errors with the articulated shape model, outperforming linear scaling.
• The model struggled to accurately capture torsional measures like femoral anteversion and tibial torsion.

Conclusions:

• The articulated shape model provides a fast and accurate method for predicting pediatric lower limb bone geometry.
• This approach is superior to linear scaling for general geometric prediction in pediatric musculoskeletal modeling.
• Further refinement is needed to improve the capture of torsional bone characteristics.