Subject-specific trunk segmental masses prediction for musculoskeletal models using artificial neural networks

Area of Science:

• Biomechanics
• Human Movement Analysis
• Machine Learning Applications

Background:

• Accurate body segment parameters are essential for musculoskeletal (MSK) modeling of human movement and joint forces.
• Current methods for predicting segment mass lack generalizability and sensitivity to diverse body shapes.
• Advancements in machine learning offer potential for improved prediction accuracy.

Purpose of the Study:

• To develop and validate a novel artificial neural network (ANN) based method for computing subject-specific trunk segment mass and center of mass (CoM).
• To utilize only anthropometric measurements as input for accurate predictions.
• To enhance the generalizability and sensitivity of segment mass prediction in MSK modeling.

Main Methods:

• Developed and trained two ANNs: ANN1 for body shape prediction and ANN2 for tissue mass prediction, using anthropometric measurements.
• Validated ANN1 on 279 subjects (max deviation 28 mm) and ANN2 on 223 subjects (mean error < 0.51% across body segments).
• Calculated trunk segmental mass for two volunteers using predicted body shape and tissue mass, comparing with experimental data.

Main Results:

• The body shape ANN (ANN1) demonstrated a maximum deviation of 28 mm across 279 subjects.
• The tissue mass ANN (ANN2) achieved a mean error of less than 0.51% for head, trunk, legs, and arms compared to ground truth.
• Calculated trunk segment mass for volunteers showed similar trends and magnitudes to experimental data.

Conclusions:

• The proposed ANN-based method provides an effective and convenient tool for predicting subject-specific trunk mass.
• This approach addresses limitations of existing methods in generalizability and sensitivity to body shapes.
• Accurate trunk mass prediction using anthropometric data can improve the fidelity of MSK models.