Separation of Neural Drives to Muscles from Transferred Polyfunctional Nerves using Implanted Micro-electrode Arrays

TL;DR

This work introduces a novel biointerface that combines TMR surgery of polyvalent nerves with a high-density micro-electrode array implanted at a single site within a reinnervated muscle, enabling the extraction of multiple neural commands within a single reinnervated muscle, eliminating the need for surgical nerve division.

Abstract

Following limb amputation, neural signals for limb functions persist in the residual peripheral nerves. Targeted muscle reinnervation (TMR) allows to redirected these signals into spare muscles to recover the neural information through electromyography (EMG). However, a significant challenge arises in separating distinct neural commands redirected from the transferred nerves to the muscles. Disentangling overlapping signals from EMG recordings remains complex, as they can contain mixed neural information that complicates limb function interpretation. To address this challenge, Regenerative Peripheral Nerve Interfaces (RPNIs) surgically partition the nerve into individual fascicles that reinnervate specific muscle grafts, isolating distinct neural sources for more precise control and interpretation of EMG signals. We introduce a novel biointerface that combines TMR surgery of polyvalent nerves with a high-density micro-electrode array implanted at a single site within a reinnervated muscle. Instead of surgically identifying distinct nerve fascicles, our approach separates all neural signals that are re-directed into a single muscle, using the high spatio-temporal selectivity of the micro-electrode array and mathematical source separation methods. We recorded EMG signals from four reinnervated muscles while volunteers performed phantom limb tasks. The decomposition of these signals into motor unit activity revealed distinct clusters of motor neurons associated with diverse functional tasks. Notably, our method enabled the extraction of multiple neural commands within a single reinnervated muscle, eliminating the need for surgical nerve division. This approach not only has the potential of enhancing prosthesis control but also uncovers mechanisms of motor neuron synergies following TMR, providing valuable insights into how the central nervous system encodes movement after reinnervation.

Figures (12)

• Figure 1: Biointerface based on Targeted Muscle Reinnervation (TMR) of polyfascicular nerve and the use of a single micro-electrode arrays for recording and decoding. On the left, a glenohumeral patient has undergone TMR nerve transfer surgery: a polyfascicular nerve that previously innervated multiple upper-limb muscles (dotted purple line) was transferred into a spare target muscle (continuous light purple line). The participant is asked to perform different tasks with his phantom limb (e.g., index finger extension) while the intramuscular activity of the reinnervated muscle is recorded using a 40-channel micro-electrode array muceli2022blind. The 40 EMG channels are distributed along 2 cm and each electrode has a diameter of 140µm (for comparison, muscle fibers have diameters in the range 10-100µm). The participant had real-time visual feedback of the median EMG activity (purple signal) that had to be modulated to match a target trapezoidal profile displayed on the screen. This resulted in an isometric contraction of the reinnervated muscle. On the right, a schematic provides insight into the reinnervation following TMR surgery. The axons of the re-routed neurons innervate fibers of the reinnervated muscles creating a heterogeneous distribution of motor units. The activity of motor units is recorded in-vivo by the intramuscular array. From the intramuscular recordings, the individual activity of the motor units can be extracted by blind-source separation methods. In this work, we hypothesize that the heterogeneous distribution of motor units corresponds to a functional organization of motor units. Thus, clusters of motor units associated with different tasks of the phantom limb can be extracted. Note that the neural structures, muscle fibers and micro-array are not represented with an accurate scaling to improve clarity.
• Figure 2: a, A micro-array of EMG electrodes, containing 40 channels (ch), is implanted into each reinnervated muscle examined to record high-density intramuscular EMG activity. b, In an exemplary task repetition, participant P3 isometrically contracts muscle TMR4 (see Tab. \ref{['tab:patient']}) while performing index finger extension with his phantom limb. A bipolar signal was derived from channels of the micro-electrode array that resulted in the maximum amplitude and was used as visual feedback of reinnervated muscle activity (purple signal). The participant modulated the muscle contraction to match the target EMG activity (dotted black line; trapezoid contraction up to a percentage of task-specific MVC). Five motor units (MUs) were reliably decomposed from the intramuscular recordings in this example. The smoothed discharges obtained by low-pass filtering the instantaneous discharge rate of each motor unit with a Hanning window of 400 ms are shown. Instantaneous discharges of active motor units are depicted with vertical lines each indicating a MU discharge at a given time instant. The smoothed and instantaneous discharges of motor units have the same color in the two plots. c, EMG voltages for active motor units on different channels of the micro-array. The discharge times of motor units is indicated with a color-coded number on top of each EMG signal. A channel records a MUAP with an amplitude that depends on the position of the detection site with respect to the fibers innervated by the motor unit. For this reason, the MUAP waveform differs across channels. d, A 2D-image of the average (across all firing instances) MUAP distribution along the 40 channels is shown for some exemplary motor units. Motor units had different morphology, as indicated by the potentials spanning fewer or almost all channels of the micro-array. e, Average MUAP of the motor units on the channel where the MUAP had maximum peak-to-peak amplitude. f, Distribution of intervals between motor unit discharges for the same motor units depicted in the panels above using 1-ms bin size.
• Figure 3: For each reinnervated muscle, the diagrams at the top detail the relation between tasks in terms of the number of shared motor units (MUs). The bar plots report the number of shared and task-specific motor units. Only MUs that could be accurately decomposed and were active in $\geq 70\%$ of task repetitions were considered.
• Figure 4: Tracking of motor units for tasks with significant time intervals between repetitions. We show an exemplary task, Pinky Flexion, performed by participant P2 at the beginning (T1) and the end (T2) of the experimental session. a, EMG recordings at task repetitions T1 and T2 were decomposed separately to identify motor units. Among the five motor units recruited in T1, 3 could be tracked in T2. The normalised MUAP of the matched motor units is shown (black and red dotted line) and the goodness of the fit between the two is quantified by the coefficient of determination $\rho$. Inb, and c, the distribution of motor units potential during T1 and T2 is shown, respectively. The dotted lines indicate the channel (Ch) at which the MUAPs had maximum peak-to-peak amplitude. A consistent shift of two channels can be observed: in T2 the matched motor units are shifted towards channel 1, indicating a slight micro-array displacement. The other two motor units identified during T1 could not be tracked in T2, possibly due to the electrode shift.
• Figure 5: Exemplary data of common synaptic input analysis for Intrinsic (TMR1).a, From left to right common oscillations in the smoothed motor units (MUs) discharges can be observed; the presence of common synaptic input to motor units is assessed by computing the cross-correlograms between smoothed and de-trended discharges of motor units (in the plateau part of the contraction); the maximum value of the cross-correlation within 100 ms of zero delay was used to quantify the strength of the common input and values are reported in the cross-correlation matrix; hierarchical clustering is applied to cluster motor units based on their inter-correlation; following the clustering analysis, motor units receiving a higher portion of common input (corr $\geq$ 0.5) are merged in a "high-correlation" cluster and "Low-correlation cluster" is formed with the MUs showing a low degree of synchronisation to other motor units (corr $<$ 0.5). b, For each task of the reinnervated muscles we report the proportion of task-specific and shared motor units that belonged to the high and low correlation groups.