Neuromorphic Decoding of Spinal Motor Neuron Behaviour During Natural Hand Movements for a New Generation of Wearable...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Current neural interfaces struggle to leverage the brain's efficient spike encoding for machine learning models.
• Spiking-based pattern recognition offers potential for adaptive, compact, and locally processed implementations in embedded systems.
• Spiking Neural Networks (SNNs) have not been previously applied to process in-vivo human neuronal activity.

Purpose of the Study:

• To propose a neuromorphic framework for processing human spinal motor neuron activity for movement intention recognition.
• To integrate this framework into a non-invasive interface for decoding motor neuron activity in hand muscles.
• To demonstrate the efficacy of convolutional SNNs for analyzing in-vivo human neuronal data.

Main Methods:

• Development of a convolutional Spiking Neural Network (SNN).
• Processing of activity from 467 spinal motor neurons across 5 participants.
• Analysis of neuronal activity during the execution of 10 distinct hand movements.

Main Results:

• Achieved a classification accuracy of approximately 0.95 ±0.14 for movement intention detection.
• Demonstrated high accuracy for both isometric and non-isometric hand contractions.
• Successfully processed in-vivo human spinal motor neuron activity using SNNs.

Conclusions:

• This study presents the first successful application of SNNs for processing in-vivo human neuronal activity.
• The proposed neuromorphic framework shows significant potential for highly accurate motion intent detection.
• Combining non-invasive neural interfaces with SNNs offers a promising pathway for advanced neuroprosthetic applications.