(K-413) Integrated EMG Processing Module for Bidirectional Neuroprosthesis

Individuals with lower limb loss frequently encounter challenges in navigating different terrains due to impaired sensory feedback and motor control. Our team aims to develop a bidirectional neuroprosthesis, which will restore plantar sensation and provide intuitive motor control over the prosthesis by interfacing with the remaining nerves and muscles in the residual limb. While prior work by our team has demonstrated the feasibility of restoring plantar sensation through implanted nerve cuff electrodes, it is unclear how this re-established sensory feedback will influence motor commands in the lower limb. To investigate this interaction, it is vital to develop an integrated system capable of capturing electromyographic (EMG) data from residual muscles, while simultaneously delivering electrical stimulation to the peripheral nerves to evoke sensory perception. Commercial EMG measurement systems, like the Delsys Trigno, are capable but are not portable or easily integrated with neural stimulation systems. A new external, portable EMG measurement system, the EPM (EMG Processing Module), was developed to capture EMG signals in real-time and interface with our laboratory’s neural stimulation systems. In this study, we report results from a series of comprehensive tests conducted to determine the EPM’s performance in recording accurate and reliable EMG from lower limb muscles.

The EPM is a compact, battery-operated, 3.5”×2.55” unit capable of measuring EMG signals from eight bipolar electrode pairs. It features an analog front-end with a 24-bit A/D, 32kHZ maximum sampling frequency, and ±4V maximum dynamic input range. The module is also equipped with a microcontroller and FPGA for onboard signal conditioning and feature extraction. We compared EPM performance to the Delsys Trigno both on two individuals with transtibial limb loss (LL1, LL2) and an able-bodied subject (AB1). We measured EMG from tibialis anterior (TA), medial gastrocnemius (MG), rectus femoris (RF), and bicep femoris (BF) muscles during maximum voluntary contractions (MVC), simple contractions, and at rest. During each trial, EMG was captured simultaneously using the same transcutaneous electrode ensuring consistent measurement from the exact same muscle location. A 20-450Hz Butterworth Bandpass filter was applied to all signals, and EMG values were divided by the respective gain of each system. We found the root mean square (RMS) value of the signal during a two second interval of MVC, and computed signal-to-noise ratio (SNR) by dividing the signal RMS by the RMS value of noise measured at rest. Frequency band power was obtained by analyzing the average power for 25Hz bands. Signal latency was evaluated by setting a threshold just above the noise to calculate the starting time of MVCs. Additionally, saturation and drift were characterized by the average range of the signals. We also assessed for presence of crosstalk between channels with muscle contractions of one muscle group while keeping another relaxed. 

Our findings show that the EPM performed comparably to the Delsys Trigno. Expressed as a percentage of the Delsys Trigno, the EPM averaged 88% for SNR and 99% for RMS, with detailed results for specific muscles shown in Table 1. Frequency band analyses demonstrated similar patterns between both systems (see Fig.1). The area under the curve (AUC) of the average power spectral density (PSD) across all trials was 244.6μV²/Hz and 276.6μV²/Hz for EPM and Delsys Trigno, respectively. The signals recorded by the EPM, on average, appeared 63ms ahead of the Delsys Trigno, which can be attributed to the latter’s ~61ms communication delay due to its wireless operation. Drift and crosstalk were negligible in the EPM signals and saturation was not observed. During periods of muscle relaxation, average drift was lower with the EPM (36.1μV) than with the Delsys Trigno (38.8μV). These values had standard deviations of 20.5μV and 20.7μV for EPM and Delsys Trigno respectively, indicating better EPM performance. The average noise attributed to crosstalk was 74.9μV, which was 4.6% of the MVC signal value. We found no saturation as the broadest range of values across all trials was within −12.5mV to 16.3mV, significantly smaller than the EPM’s ±4V input range.

EPM performance was equivalent to commercially available devices for the assessed metrics, which underscores its potential as an effective tool for portable EMG measurements and a crucial component in the advancement of the bidirectional prosthesis. Future research will involve the use of the EPM for collecting EMG with intramuscular electrodes in lower limb amputees who have been implanted with this technology. Furthermore, we aim to employ the EPM for EMG measurements during ambulatory tasks while participants receive sensory neural stimulation which elicits sensations in their missing foot.

This project was supported in part by the Department of Defense under Awards No. W81XWH-18-1-0321 and W81XWH-20-1-0802. This work was also supported by Wen H. Ko Summer Internship Program sponsored by the Advanced Platform Technology Center at Louis Stokes Cleveland VA Medical Center. In addition, this submission is the result of work supported with resources and the use of facilities at the Louis Stokes Cleveland VA Medical Center.