A PRISMA systematic review through time on predictive musculoskeletal simulations.

Area of Science:

• Biomechanics and computational modeling.
• Systematic review of predictive musculoskeletal simulations.

Background:

• Inconsistent nomenclature for predictive movement in musculoskeletal systems necessitates clear definitions.
• Two primary prediction methods exist: neural models (muscle-reflex, central pattern generators) and optimization (optimal control, deep reinforcement learning).

Observation:

• The majority of reviewed works utilize optimization-based prediction methods.
• Six classes of cost function terms are identified, often combined linearly.
• Key musculoskeletal modeling components include muscle models, ground contact, and personalization.

Findings:

• Optimization methods are prevalent, with identified accuracies, simulation times, and limitations.
• Key historical works and software (OpenSim, SCONE) are highlighted.
• Experimental data enhances predictive movement simulations.

Implications:

• Standardized quantitative validation is crucial for comparing predictive simulation methods.
• Open-sourcing code facilitates reproducible research and advancement in the field.
• Understanding predictive simulations aids in analyzing movements, pathologies, and assistive devices.