Targeting Transcutaneous Spinal Cord Stimulation Using a Supervised Machine Learning Approach Based on Mechanomyography

Sensors, 2024 · DOI: 10.3390/s24020634 · Published: January 19, 2024

Simple Explanation

This paper explores using mechanomyography (MMG) and machine learning to simplify the process of tuning transcutaneous spinal cord stimulation (tSCS). tSCS is a promising therapy for individuals with spinal cord injuries and multiple sclerosis patients, but the calibration procedure is complex. The study aims to replace electromyography (EMG) with MMG, which uses accelerometers to assess muscle activity, making the process easier and more accessible. The researchers implemented a supervised machine learning classification approach to classify acceleration data into no activity and muscular/reflex responses. This was done using EMG responses as ground truth. The acceleration-based calibration procedure achieved a mean accuracy of up to 87% relative to the classical EMG approach. The study concludes that MMG has the potential to make the tuning of tSCS feasible in clinical practice and even in home use. This simplification could improve accessibility and reduce the need for expert knowledge.

Study Duration

Not specified

Participants

11 healthy subjects and 11 MS patients

Evidence Level

Not specified

Key Findings

1
The acceleration-based calibration procedure achieved a mean accuracy of up to 87% relative to the classical EMG approach as ground truth on a combined cohort of 11 healthy subjects and 11 patients.
2
The identified current amplitude for the tSCS therapy was in 85%, comparable to the EMG-based ground truth.
3
In healthy subjects, where both therapy current and position have been identified, 91% of the outcome matched well with the EMG approach.

Research Summary

The study investigates the use of mechanomyography (MMG) and machine learning (ML) to simplify transcutaneous spinal cord stimulation (tSCS) calibration. It proposes using accelerometers to measure muscle activity instead of electromyography (EMG). A supervised ML approach was used to classify acceleration data into activity or no activity, using EMG as the ground truth. The method achieved up to 87% accuracy compared to the traditional EMG approach. The findings suggest that MMG, combined with machine learning, can make tSCS tuning more accessible and feasible for clinical and home use, reducing the reliance on expert knowledge and complex setups.

Practical Implications

Simplified tSCS Calibration

MMG combined with machine learning can simplify the tSCS calibration process, potentially making it more accessible and practical for clinical and home use.

Reduced Reliance on Expertise

The automated approach reduces the need for expert knowledge in sensor placement and signal interpretation, which can broaden the availability of tSCS therapy.

Cost-Effectiveness and Sustainability

The use of IMUs instead of disposable EMG electrodes can decrease long-term costs and improve sustainability.

Study Limitations

1
The accuracy of the classifiers could be improved, particularly in distinguishing between reflex responses and direct muscular responses.
2
Crosstalk in the acceleration signals due to joint movement or contraction of other leg muscles could affect the accuracy of the MMG data.
3
The limited quantity of patient data may affect the generalizability of the machine learning models, suggesting a need for larger datasets from specific patient groups.

Your Feedback

Was this summary helpful?

Back to Neurology