Related Experiment Videos
Evaluation of artificial neural network modelling to predict torso muscle activity
1Center for Ergonomics, Industrial and Operations Engineering, University of Michigan, Ann Arbor 48109-2117, USA.
Ergonomics
|December 1, 1996
Summary
This study validates an artificial neural network (ANN) model for predicting lumbar muscle activity during exertion. The ANN model accurately predicted muscle activation within 3%, demonstrating its reliability for generalized recruitment patterns.
Area of Science:
- Biomechanics
- Computational modeling
- Neuroscience
Background:
- Direct measurement of in vivo muscle forces is complex, hindering predictive model validation.
- An artificial neural network (ANN) model was previously developed to estimate lumbar muscle activity during static exertion.
Purpose of the Study:
- To evaluate the predictive accuracy of a previously developed ANN model for lumbar muscle activity.
- To compare ANN model predictions with experimental electromyography (EMG) data.
Main Methods:
- Utilized response surface comparisons and composite statistical tests.
- Compared ANN model output with multiple EMG experimental datasets.
- Evaluated model performance across various flexion/extension and lateroflexion loadings.
Main Results:
- ANN-predicted activation levels were accurate within 3% across diverse experimental conditions.
- Demonstrated high consistency in averaged muscle activity measurements.
- Substantiated the ANN model's ability to predict generalized muscle recruitment patterns.
Conclusions:
- The ANN model accurately predicts lumbar muscle activity during static exertion.
- Multiple comparison methods offer superior evaluation of model behavior and prediction accuracy compared to single criteria.
- The findings support the use of ANN models for understanding muscle recruitment patterns in biomechanical research.