Related Experiment Video
Updated: Aug 5, 2026
06:52
Deep-Learning Based Multi-Joint Synchronous Tracking for Objective Quantification of Hindlimb Locomotor Kinematics in Rats
Published on: April 3, 2026
OpenSim-Umberger-Based Metabolic Power Stratification During the Sit-to-Walk Transition Using Interpretable Ensemble
Wanli Zang1, Jiarong Wu1, Jun Wu1
1School of Physical Education, Soochow University, Suzhou 215021, China.
Bioengineering (Basel, Switzerland)
|July 28, 2026
Summary
This study introduces a new method using machine learning to estimate metabolic cost during sit-to-walk (STW) transitions, improving analysis of movement efficiency.
Area of Science:
- Biomechanics
- Metabolic Physiology
- Machine Learning
Background:
- Quantifying metabolic cost in short movements is difficult due to limitations in traditional measurement techniques.
- The sit-to-walk (STW) transition is a critical functional movement with poorly understood metabolic demands.
- Interpretable machine learning offers a potential solution for analyzing complex biomechanical data.
Purpose of the Study:
- To investigate the feasibility of stratifying model-derived metabolic cost during the STW transition using interpretable ensemble learning.
- To identify key biomechanical features that predict different levels of metabolic cost during STW.
- To develop and validate a computational workflow for assessing metabolic cost in transitional movements.
Main Methods:
- Utilized 3D kinematics, ground reaction forces, and EMG data from 49 healthy adults performing the STW transition.
- Employed OpenSim and Umberger metabolic models to estimate metabolic power.
- Extracted biomechanical features and trained seven classifiers, fusing their outputs using TOPSIS.
- Applied SHapley Additive exPlanations (SHAP) for feature attribution.
Main Results:
- The fused ensemble model achieved an AUC of 0.870, F1 score of 0.703, and accuracy of 0.705 on an independent test set.
- The model demonstrated better discrimination for low and high metabolic cost levels compared to the medium level.
- SHAP analysis identified force-related variables and knee kinematics (variability, amplitude) as key predictors of metabolic cost.
Conclusions:
- A model-derived, interpretable machine learning approach can effectively stratify metabolic cost during the STW transition.
- This method offers a promising way to assess metabolic cost beyond simple task performance metrics.
- Further validation in larger and clinical populations is warranted to confirm generalizability.
