Researcher(s)
- Connor Spinelli, Biomedical Engineering, University of Delaware
Faculty Mentor(s)
- Fabrizio Sergi, Biomedical Engineering, University of Delaware
Abstract
The central nervous system’s task-dependent modulation of motor responses, specifically the supraspinal pathway reticulospinal tract (RST), is fundamental to human motor control. When a muscle is rapidly stretched, an involuntary stretch reflex is triggered, involving spinal and supraspinal pathways. Long-latency responses (LLR) are a critical component of this reflex, as they are adjustable based on a task’s goals; this adaptability is key for post-stroke motor rehabilitation. To investigate the RST’s role in modulating LLR, we rely on accurate measurements of muscle activity using surface electromyography (sEMG). sEMG signals contain noise, necessitating a three-step pre-processing pipeline: filtration, normalization, and averaging. Since interpreting these signals is highly dependent on these steps, our study aimed to validate our pre-processing approach by examining how filtering parameters influence the statistical significance of LLR amplitudes across different task instructions.
We recorded sEMG signals from the wrist flexor and extensor muscles (flexor carpi radialis and extensor carpi ulnaris) of 10 participants. They completed a wrist perturbation task under three distinct instructions known to modulate LLR amplitudes: “Yield” to the perturbation, “Resist” the perturbation, or “Slow” (yielding to a slower perturbation). To assess the impact of pre-processing, we compared the results from eight different filter setups. These setups varied by filter design (Butterworth or Blackman window), passband (10-350 Hz or 20-250 Hz), and the inclusion or exclusion of linear enveloping.
A mixed-model analysis revealed that the statistically significant differences in LLR amplitudes between the task instructions remained consistent across all filter setups. Our findings show that the significance of our results was not influenced by the specific filtering parameters chosen. This consistency validates our methodology, confirming that our conclusions are robust and not artifacts of our pre-processing choices. Our results support a broader investigation into RST function and suggest that while sEMG filtering is important, the specific techniques chosen may not critically affect the statistical significance of results in similar studies.