From Neuromuscular Activation to End-point Locomotion: An Artificial Neural Network-based Technique for Neural Prostheses

J Biomech, 2009 · DOI: 10.1016/j.jbiomech.2009.03.030 · Published: May 29, 2009

Simple Explanation

The study investigates if an Artificial Neural Network model can use neuromuscular activation to predict gait parameters in individuals with impaired spinal cords for neural prostheses. The study recruited 12 individuals with Spina Bifida, collecting neuromuscular activation and gait parameters during overground walking. The results suggest that engaging neuromuscular activity as intrinsic feedback can achieve more precise control of complex neural prostheses during locomotion.

Study Duration

Not specified

Participants

12 persons with Spina Bifida

Evidence Level

Not specified

Key Findings

1
The ANN-based technique achieved highly accurate predictions for altered end-point locomotion (R-values of 0.92 – 0.97).
2
The ANN-based technique can provide up to 80% more accurate prediction compared to traditional robust regression.
3
ANN-based prediction schemes consistently outperformed regression-based techniques with considerably better accuracies.

Research Summary

The study developed an ANN-based technique to predict end-point limb motions using neuromuscular activity feedback from individuals with interrupted spinal cords. Experimental results confirmed the high prediction accuracy (R-values of 0.92 – 0.97) of the proposed technique. The findings suggest that neuromuscular information can successfully predict extrinsic end-point locomotion by applying the proposed ANN-based technique.

Practical Implications

Neural Prosthesis Interface

ANN-based technique with neuromuscular activity feedback could form the basis of neural prosthesis interface for paralyzed individuals.

Adaptive Controller Development

Future work will focus on developing a self-organizing and adaptive controller using low-power hardware-software co-design techniques.

Improved Locomotion Prediction

Intrinsic neuromuscular information recorded through EMG sensors can successfully predict extrinsic end-point locomotion.

Study Limitations

1
The study focused on individuals with Spina Bifida, limiting generalizability to other populations with paralysis.
2
The exact location of the point of diminishing return for the number of hidden neurons can vary depending on the input data and other ANN network parameters.
3
The study used a specific ANN architecture (three-layer feed-forward network), and other architectures may yield different results.

Your Feedback

Was this summary helpful?

Back to Rehabilitation