Decoding Overlapping Lower-Limb Afferent Pathways from Human Epidural Spinal Recordings

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Adaptive neuroprostheses require real-time sensory feedback for effective motor function restoration after neurological injury.
• Decoding distinct afferent nerve activity from overlapping epidural signals, especially in the lumbosacral region, is a significant challenge.

Purpose of the Study:

• To determine if clinical-grade lumbosacral epidural paddle arrays can decode distinct afferent activity from overlapping signals.
• To investigate the feasibility of using these signals for closed-loop neuromodulation.

Main Methods:

• Utilized a 32-contact lumbosacral epidural paddle array and peripheral nerve stimulation.
• Constructed a 42-dimensional feature space including amplitude, field geometry, and waveform morphology.
• Employed a support vector machine classifier and Shapley Additive Explanations for decoding and feature importance analysis.

Main Results:

• Successfully decoded four distinct afferent classes (left and right common fibular nerve and tibial nerve) with 90.9% median accuracy.
• Identified that contact-level voltage patterns and temporal waveform complexity were key decoding features.
• Demonstrated that afferent-specific signatures were detectable even at sub-motor threshold stimulation intensities.

Conclusions:

• Clinical-grade epidural arrays can capture sufficient spatiotemporal information to decode overlapping lumbosacral afferent inputs.
• This approach offers a viable, less invasive pathway for developing interpretable closed-loop neuromodulation systems.
• Potential for rapid deployment in restoring motor function following neurological injury without requiring surgical nerve interfaces.