Performance of Two Different Targeted Muscle Reinnervation Approaches for Improving Myoelectric Prosthetic Control

Targeted Muscle Reinnervation (TMR) is a surgical approach that can produce a neural-machine interface to provide additional electromyography (EMG) signals in intuitive control of myoelectric prostheses for high-level limb amputees. Clinically, TMR interface can be built by two types of surgical protocols. The first surgical protocol is to transfer a residual nerve to a denervated targeted muscle using a nerve-to-muscle suture and another is to anastomose a residual nerve to a targeted nerve via a nerve-to-nerve suture. Currently, it still remains unknown which surgical protocol would be more suitable for their clinical applications. In this study, a comparative investigation of the two TMR protocols was conducted by using animals and their performance in reconstructing EMG signals was evaluated. For nerve-to-muscle animal model, the proximal end of ulnar nerve was implanted onto denervated biceps brachii muscle and for nerve-to-nerve model, the proximal end of ulnar nerve was anastomosed to the distal end of musculocutaneous nerve. Post-surgery EMG signals were collected from all the TMR animals. Our results showed that the amplitudes of EMG signals gradually increased for the animals in the two TMR protocols over time, with an obvious difference between nerve-to-muscle and nerve-to-nerve animals. The signal-to-noise ratio and the centroid frequency of EMG signals in nerve-to-muscle animals were notably higher than those in nerve-to-muscle animals. These superior characteristics of the postoperative EMG signals demonstrated that nerve-to-nerve surgical protocol would be better than nerve-to-muscle in reconstructing EMG signals. The findings of this animal study suggest that both the TMR surgical protocols could be useful in providing additional EMG signals for myoelectric control, while the nerve-to-nerve suture would outperform the nerve-to-muscle suture in the quality and the appearance of the reconstructed extra EMG signals.

SECTION I. Introduction

Millions of people worldwide are suffering from limb loss due to accidents, industrial injuries, diseases, and more. Losing limbs limit their daily activities, and wearing functional prostheses becomes an important choice for amputees [1], [2]. At present, myoelectric prostheses are the most popular solution to offset partial lost limb movement function [3], [4]. However, for above-elbow and above-knee amputees, the deficiency of EMG signals from residual limbs makes myoelectric prostheses be difficult to achieve intuitive multifunctional prosthetic control, resulting in a high ratio of myoelectric prostheses abandonment [5], [6], [7].

Targeted muscle reinnervation (TMR) has been developed to enhance the control of multifunctional myoelectric prostheses for high-level amputees [8], [9], [10], [11]. Unlike traditional myoelectric control, which relies on a small number of residual muscle sites, TMR provides more signal sources, enabling greater dexterity and natural movement. Currently, TMR has been widely adopted, helping many amputees enhance their control over myoelectric prostheses and reducing the prosthesis abandonment rate. Particularly noteworthy is that, when combined with pattern recognition, it enhances decoding accuracy, improves intuitive control, and enables more natural and versatile prosthetic movements [5]. Moreover, when combined with osseointegration, it eliminates the need for socket-based interfaces, improving comfort, stability, and long-term usability [12]. These advancements significantly enhance prosthetic functionality, ultimately improving the quality of life for amputees.

TMR surgery involves transferring the residual nerves to targeted muscles [13]. Once those targeted muscles are reinnervated by implanted nerves, they serve as a bio-amplifier characterizing motor intention [14]. A previous study firstly reported a male subject with bilateral shoulder disarticulation who accepted a TMR surgery: four brachial plexus nerves were respectively implanted onto different areas of the pectoralis major [8]. Six months after surgery, strong and independent contractions were observed, and EMG signals could be detected on the surface of his chest when he attempted movements of his missing upper limb. Objective assessments demonstrated a two-fold increase in the number of blocks relocated during the box and blocks test, alongside a 26% enhancement in task execution speed during the clothes pin relocation test. Similarly, another study reported a subject with unilateral transfemoral amputation who accepted a TMR surgery, in which a common fibular nerve and tibial nerve were respectively transferred to long head of biceps femoris and semitendinosus [15]. As a results, after a few months, there were strong, independent contractions and EMG signals that could be sensed on the surface of his dorsal thigh when he wanted to execute movements about his lost knee and ankle. A study reported three transhumeral amputees who accepted TMR surgeries. In their surgical procedures, stump median and radial nerves were anastomosed to the nerve of targeted muscles. The post-surgery after six months showed that, there were strong, independent contractions, and EMG signals that could be sensed from four different myoelectric sites: hand-closing from the medial head of the biceps, which was reinnervated by the median nerve; elbow flexion from the lateral head of the biceps; hand-opening from the reinnervated brachialis muscle; and elbow extension from the triceps [16].

Through the analysis of the specific surgical procedures of the aforementioned TMR cases, it can be observed that there are two approaches to perform a TMR surgery: nerve-to-muscle suture involves transferring a nerve to a targeted muscle [8], [17], and nerve-to-nerve suture involves anastomosing a nerve to the nerve supplying the targeted muscle [16], [18], [19]. Nerve-to-nerve suture is a common clinical approach for peripheral nerve function repair [20]. With the introduction of TMR, researchers further developed the nerve-to-muscle suture to reconstruct EMG sites and optimize motor function recovery [8]. We believe that the nerve-to-nerve suture has greater advantages, as it directly connects the two nerve stumps through epineurial suturing, making full use of the original donor nerve pathways to achieve rapid reconstruction of bands of Büngner, thereby promoting nerve regeneration. In contrast, the nerve-to-muscle suture involves direct suturing of the nerve to the muscle, lacking the support of the original nerve channels [21], [22]. Additionally, the presence of structures such as the muscle epimysium may create barriers, potentially hindering nerve regeneration efficiency and outcomes. However, there is still a lack of quantitative studies comparing these two approaches. For instance, what are the differences in EMG signal characteristics between nerve-to-nerve and nerve-to-muscle sutures? How do the two methods affect the morphology of the target muscle? To address these questions, this study would employ quantitative analysis to systematically evaluate the differences between nerve-to-muscle and nerve-to-nerve sutures.

Thus, we conducted a comparative investigation of the nerve-to-muscle and nerve-to-nerve approaches in reconstructed EMG signals, in which rats were used and divided into two groups to build TMR models, respectively. One group of rats undergone the nerve-to-muscle suture that transferred the ulnar nerve (UN) onto the homolateral biceps brachii muscle (BBM), and another group undergone the nerve-to-nerve suture that anastomosed the UN to the distal end of the musculocutaneous nerve (MCN) supplying the homolateral BBM. The reconstructed EMG signals from both TMR models were examined by using tetanic contractility force (TCF), intramuscular nerve distribution (IND), and muscle atrophy of those TMR BMMs. This study was expected to clarify the differences between nerve-to-muscle and nerve-to-nerve approaches in reconstructing EMG signals.

SECTION II. Methods

A. Surgical Procedures

In this study, 14 specific pathogen-free (SPF) Sprague Dawley adult male rats, weighing approximately 200 grams each, were randomly divided into nerve-to-muscle and nerve-to-nerve groups, with equal numbers in each group. The rats were anesthetized by isoflurane (5% and 2% isoflurane for induction and maintenance, respectively) and then placed on a heating plate to maintain the body temperature (37 °C). The skin covering the BBM was shaved off and sterilized with 5% povidone iodine solution before surgery. Those experimental procedures of this study were assessed and approved by the Committee on the Use and Care of Animals at the Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences (SIAT).

Figure 1 represents the procedures for the nerve-to-muscle and nerve-to-nerve surgeries conducted in this study. For nerve-to-muscle group, incisions of approximately 3 cm were made parallel to the forelimbs to access the BBMs and brachial plexus. The right UNs and MCNs of rats in the nerve-to-muscle group were transected with a surgical knife. Then, the proximal end of the UN was implanted onto the BBM’s belly using a surgical suture needle with thread (#11-0). Sequentially, the ends (both distal and proximal) of the MCN and the distal end of the UN were ligated. It should be noted that ligating the ends of the nerves was necessary to prevent them from connecting with surrounding nerves or muscles that could lead to unwanted phenomena, which could introduce additional variables affecting the function of the target nerve and muscle. For the nerve-to-nerve group, after transecting the UN and MCN same to nerve-to-muscle surgery, a proximal end of UN was anastomosed to a distal end of MCN via the epineurium, with other two ends ligated.

Fig. 1.

Schematic diagram of the surgical paradigm. (A) The self-designed wireless EMG signals recorder is connected to the model through a skull connector, and the rectangular boxes indicate the surgical sites of the forelimbs on both sides. (B) Enlarged view of the content within the rectangular box in (A), and the left image represents the normal anatomical structure, where the musculocutaneous nerve (MCN) shown in blue innervates the biceps brachii muscle (BBM), with the ulnar nerve highlighted in yellow. The middle image represents nerve-to-muscle suture, where the ulnar nerve (UN) is implanted into the denervated BBM. The right image represents nerve-to-nerve suture, where an end-to-end anastomosis is made between the UN and the MCN. (C-D) The surgical scenes during nerve-to-muscle and nerve-to-nerve suture, respectively.

Additionally, for recording EMG signals from a running rat, a skull connector would be immobilized onto rat’s skull [23]. The use of a skull connector for EMG wire placement offers several advantages: firstly, minimizing mechanical strain on EMG wires. Secondly, improving long-term recording stability. This setup ensures stable and reliable signal acquisition, particularly in freely moving animals. In details, firstly to expose parietal skull and drill a few screws into the holes without piercing the dura mater. Then skull connector linked with five stainless wires covered with Teflon insulated coat were adopted, and all the ends were exposed 4-milimeter length, a pair of wires were subcutaneously pulled and implanted onto the right BBM (TMR BBM), another a pair of wires were subcutaneously pulled and implanted onto the left BBM (control BBM), and the fifth wire served as the ground electrode fixed onto the screw. Finally, ultra-violet glue was used to fix the skull connector and skin incisions were sutured. After the surgeries, penicillin sodium (Dose: 100000 IU/rat) and meloxicam (Dose: 2 mg/kg) were administrated via intraperitoneal injection to avoid infection and pain for continuous three days, respectively. Afterwards, all nerve-to-muscle and nerve-to-nerve rats were kept in an SPF room with controlled temperature ($22\sim 26^{\circ }$ C), humidity ($40\sim 70$ %), and a day/night cycle (12:12 light/dark) and given ad libitum access to food and water for three months.

B. EMG Signal Recording and Preprocessing

This study primarily aims to compare whether there are differences in the EMG signals generated by the two TMR approaches. Naturally, this involves recording and analyzing the quality and characteristics of the EMG signals produced by both methods. Animal models from both groups were scheduled to walk at a speed of 5 m/min on a treadmill, and a customized wireless electrophysiological collecting equipment (NES16A01, SIAT, Shenzhen, China) was used to record EMG signals, with a sampling rate of 2000 Hz. EMG signals were recorded at 7-day intervals, up to 28 days post-surgery, for a total of four sessions. To preserve the EMG signal components of interest while minimizing the impact of low-frequency noise and high-frequency artifacts, bandpass filtering was then applied using a 3th order Butterworth filter with a 20 Hz to 350 Hz pass band, implemented using BUTTER and FILTER functions in MATLAB R2022a (MathWorks, USA) [24]. To better understand the features of EMG signals recorded from TMR animals, we calculated the signal-to-noise ratio (SNR, formula 1), normalized EMG (formula 2), and centroid frequency (CF, formula 3).

$$\begin{equation*} \mathrm { SNR}={10}\ast \log \left [{{ \frac {\mathrm {RMS}\left ({{ \mathrm {EMG} }}\right)}{\mathrm {RMS}\left ({{ \mathrm {background ~noise} }}\right)} }}\right ] \tag {1}\end{equation*}$$ in which, RMS() is a function in Matlab, used to calculate root-mean-square value a vector, and log indicates a function of log base ten in Matlab. $$\begin{equation*} \mathrm {normalised~EMG}=\frac {\mathrm {rms(TMR~EMG)}}{\mathrm {rms(control ~EMG)}} \tag {2}\end{equation*}$$ in which, rms() is a function in Matlab. $$\begin{equation*} \mathrm {CF}=\frac {\sum \nolimits _{\mathrm {i=1}}^{\mathrm {n}} {\mathrm {f}_{\mathrm {i}}\mathrm {.\ast }\mathrm {p}_{\mathrm {i}}} }{\sum \nolimits _{\mathrm {i=1}}^{\mathrm {n}} \mathrm {p}_{\mathrm {i}} } \tag {3}\end{equation*}$$ in which, f and p are the outputs of the FFT function in Matlab, f is the frequency trains, and p is the amplitude of each f point, n is the length of EMG signal.

SNR is one of the time-domain metrics for evaluating EMG signals quality and can effectively compare the quality of EMG signals reconstructed by the two surgical approaches. Normalized EMG is also a time-domain metric for assessing EMG signals quality, where the TMR EMG signals are standardized using the control EMG signals, fully accounting for inter-individual differences in EMG signals. The CF is a frequency-domain indicator of the EMG signals. Although CF is not typically used to compare the quality of EMG signals, in this study, the authors aimed to use the CF to compare the characteristics of EMG signals generated by the two approaches. Additionally, we sought to determine whether these differences could reflect the quality of the EMG signals.

C. Tetanic and Single Contractility

Contractility properties serve as functional indicators that effectively reflect the extent of neural regeneration on a targeted muscle. Better neural regeneration corresponds to stronger innervation, manifested by the ability of targeted muscles to generate greater force under neural impulses, and vice versa. Three months after the surgery, all models underwent tests to measure TCF and maximal single contractility force (MSCF). Prior to these tests, the models were deeply anesthetized by intraperitoneal injection of sodium pentobarbital (dose: 60 mg/kg body weight, density: 10 mg/ml). The bilateral BBMs and their nerves were exposed. An inelastic string was then tied to the distal tendon of the muscle, and the other end of the string was attached to a transducer which was connected to the recording input channel of a MedEasy system (U/8C502, Nanjing, China). Customized stimulating hook electrodes were attached to the nerve, and the other ends were connected to the stimulus output channel of the MedEasy system. The following stimulation parameters were used for the study: a pulse width of 0.2 ms, and an amplitude range of 0.4-2 V with an increment of 0.4 V for the MSCF tests. For TCF tests, the stimulation pulse width was 0.2 ms and the frequency was increased from 11 to 51 Hz with a step of 10 Hz, and the stimulus voltage equaled 2 V [25], [26]. Under electrical stimulation above 41 Hz, the muscle enters a tetanic state, rapidly contracting to its maximum force and maintaining this state until the stimulation ceases or muscle fatigue occurs. The response duration (DR) refers to the time difference, measured in milliseconds (ms), between the onset of the tetanic state and the point at which the increase in TCF begins to slow down. This metric is typically used to reflect muscle cells’ coordination and fatigue resistance.

D. Muscle Atrophy

Skeletal muscle typically experiences atrophy when there is a lack of adequate nutrition from the neuron, resulting in a decrease in both volume and mass. Thus, morphology serves as the foundation for functional recovery, and good morphology often indicates good functionality. Thus, after completing the overall neuromuscular motor function assessment through in vivo experiments, the muscle atrophy status was further evaluated by fully isolating the bilateral BBMs and removing the tissues covering their surface to ensure that the tendons and muscles remained intact. Each BBM mass was weighted, and the preservation ratio (PR, formula 4) was designed to assess muscle atrophy for nerve-to-nerve and nerve-to-muscle models.

$$\begin{equation*} PR=\frac {\mathrm {TMR}~BBM~mass}{\mathrm {control}~BBM~mass} \ast 100\% \tag {4}\end{equation*}$$

E. Histological Staining

After a TMR surgery, BBM lacks neural nutrition from neuron body, and went through axons degeneration and muscle atrophy. To investigate them, models in nerve-to-nerve and nerve-to-muscle groups were euthanized with an intraperitoneal injection of sodium pentobarbital (dose: 800 mg/kg body mass, density: 240 mg/ml), and their targeted muscles were collected for histological staining. The Sihler’s nerve staining technique was used to reflect mapping of the whole nerve distribution patterns of a musculature. The staining technique followed a standard operation protocol, and made musculature transparent and nerve in purple. For myocytes’ visualization, standard paraffin embedding and segmentation with a thickness of $8~\mu$ m were necessary, then masson trichrome staining (hematoxylin, eosin, and light green) was used to stain muscle cells, and the myocytes, collagenous fibers, and nuclei were stained in red, green, and purple/black, respectively.

F. Statistical Analysis

The ANOVA and independent samples t-test statistical methods were employed to compare the additional EMG signal, TCF, MSCF, PR, and myocytes’ size of the targeted muscle between nerve-to-nerve and nerve-to-muscle groups. Prior to conducting these analyses, normality tests were performed to assess the distribution of each dataset. Data that followed a normal distribution were analyzed using ANOVA or independent samples t-tests. For data that did not meet the normality assumption, the Mann-Whitney U test was used as a non-parametric alternative. It is worth noting that all the statistical results were obtained at a significance level of $p \lt 0.05$ (95% confidence level), and the results have been reported in terms of mean ± standard deviation (SD).

SECTION III. Results

In the initial days after the TMR surgeries, the models showed good recovery, with smooth fur and no signs of stress or behavioral abnormalities. The two surgeries had a time difference, with the nerve-to-muscle suture taking $42.7~\pm ~3.2$ minutes and the nerve-to-nerve suture taking $32.6~\pm ~2.1$ minutes. Figure 1C and 1D illustrate the nerve-to-muscle and nerve-to-nerve surgical procedures, respectively. Additionally, the electrodes implantation and skull connector immobilization procedures for EMG recording took approximately $64.9~\pm ~15.6$ minutes.

A. Electromyography Signal

The key determinant for the functional connection between a transferred nerve and a targeted muscle lies in the ability of the targeted muscle to generate EMG signals evoked by the nerve. In this study, EMG signals were recorded from nerve-to-muscle and nerve-to-nerve models for 28 days post-surgery. Figure 2A–​2B and 2C–​2D illustrate the EMG signals of the TMR and control BBMs of both group models, respectively, during steady treadmill locomotion. Each figure presents four traces from top to bottom, representing the EMG signals recorded on postoperative days 7, 14, 21, and 28. Based on figure 2A and 2C, it was observed that the amplitude of the EMG signals from the TMR BBMs gradually increased with the duration of rehabilitation in both groups of TMR models. For nerve-to-muscle models, the root mean square (RMS) of EMG signals generated by the TMR BBMs were $10.03~\pm ~0.44~\mu$ V on day 7 and $22.34~\pm ~0.33~\mu$ V on day 28. For nerve-to-nerve models, the RMS of EMG signals generated by the TMR BBMs were $7.5~\pm ~0.94~\mu$ V on day 7 and $28.49~\pm ~0.11~\mu$ V on day 28. Besides, when observing the EMG signals of the control BBMs, it was found that the signals in both nerve-to-muscle and nerve-to-nerve models (figure 2B and 2D) significantly ($p \lt 0.05$ ) exhibited a higher RMS ($50.09~\pm ~10.67~\mu$ V, $40.74~\pm ~15.59~\mu$ V, respectively).

Fig. 2.

EMG signals recorded from models during treadmill exercise. (A-D) The EMG signals of nerve-to-muscle and nerve-to-nerve models at 28 days post-surgery, with each row depicting EMG signals from postoperative days 7 (top trace), 14, 21, and 28, respectively. (A-B) The EMG signals of the targeted muscle reinnervation (TMR) and control biceps brachii muscles (BBMs) in nerve-to-muscle model, respectively. (C-D) The EMG signals of the TMR and control BBMs in nerve-to-nerve model, respectively. (E) The signal-to-noise ratio (SNR) of EMG signals on the TMR BBMs in nerve-to-muscle and nerve-to-nerve models. (F) The normalized EMG signals on the TMR BBMs in nerve-to-muscle and nerve-to-nerve modesl. (G) The normalized centroid frequency (CF) of the EMG signals on the TMR BBMs in nerve-to-muscle and nerve-to-nerve models. (${}^{\ast } \text {)}$ there is a significant differences at ${p} \lt 0.05$ , (${}^{\ast \ast }\text {)}$ there is a significant differences at ${p} \lt 0.01$ .

Figure 2E displayed the SNR in the TMR EMG signals of both groups, indicating that the nerve-to-muscle models had significantly lesser SNR than the nerve-to-nerve models after day 14 ($p \lt 0.05$ ). Additionally, when observing the normalized EMG signals of TMR BBMs (Figure 2F), there were significantly higher values in the nerve-to-nerve models compared to the nerve-to-muscle models ($p \lt 0.05$ ). Furthermore, there were significant differences in the CF of the TMR EMG signals between the two groups ($p \lt 0.01$ ). In details, the CF of the nerve-to-nerve models was $182.18~\pm ~9.18$ Hz, which did not differ significantly from the control EMG signals ($176.50~\pm ~5.39$ Hz), while the CF of the nerve-to-muscle models was $128.90~\pm ~8.74$ Hz. Moreover, the CF values after normalization using the CF of control EMG signals as displayed in figure 2G, indicated significant differences between the two groups on days 7, 14, 21, and 28 ($p \lt 0.05$ or $p \lt 0.01$ ).

B. Contractility

Figure 3A demonstrates different contraction patterns of targeted muscles in nerve-to-muscle and nerve-to-nerve models with increasing stimulation frequency. As the stimulation frequency increases to 11 Hz, 21 Hz, and 31 Hz, the contraction frequency of the targeted muscles becomes faster, characterized by incomplete TCF with a serrated waveform, and the force generated increases. When the stimulation frequency is further increased to 41 Hz and 51 Hz, both nerve-to-muscle and nerve-to-nerve models exhibit complete TCF. It can be observed that with further increases in stimulation frequency, the TCF of targeted muscles in both nerve-to-muscle and nerve-to-nerve models continues to increase, reaching its maximum at a stimulation frequency of 41 Hz. At stimulation frequencies (11 Hz, 21 Hz, 31 Hz, 41 Hz and 51Hz), the TCF force of the nerve-to-nerve group is significantly higher than that of the nerve-to-muscle group ($p \lt 0.05$ ).

Fig. 3.

Contractility characteristics of biceps brachii muscles (BBMs). (A) Tetanic contractility force (TCF) of the targeted muscle reinnervation (TMR) BBMs in both nerve-to-muscle and nerve-to-nerve models at increasing electrical stimulation frequency, solid and dash arrows represented. (B) Response duration (RD) of TCF. (C) Comparison of maximum single contractility force (MSCF) in the TMR BBMs at increasing electrical stimulation voltage between nerve-to-muscle and nerve-to-nerve models. (${}^{\ast } \text {)}$ there is a significant differences at ${p} \lt 0.05$ .

The RD of the TCF, defined as the time from the rapid initial rise of the stimulation to a flat portion of the complete TCF trace, is shown in Figure 3B. The RD of nerve-to-muscle BBMs is significantly higher than that of nerve-to-nerve BBMs ($p \lt 0.05$ ), indicating better consistency of neural activation of targeted muscles in nerve-to-nerve models. Additionally, for MSCF of BBMs, nerve-to-nerve suture also outperforms nerve-to-muscle suture (Figure 3C). Specifically, as the stimulation voltage increases, the MSCF of targeted muscles in the nerve-to-nerve group gradually increases and reaches its maximum at 1.6 V. In contrast, the MSCF of targeted muscles in the nerve-to-muscle group does not show enhancement and is significantly lower than that of the nerve-to-nerve group starting from a stimulation voltage of 0.8 V ($p \lt 0.05$ ).

C. Intramuscular Nerve Distribution and Muscle Physiology

At the microscopic level, figure 4A displays the staining of cross-sections of BBMs from nerve-to-muscle and nerve-to-muscle models. The left panel shows the Masson’s trichrome staining results, while the right panel depicts the contour map of muscle fibers. The muscle fibers of both groups are clearly visible, and in nerve-to-muscle BBMs, some additional green clusters can be observed, which represent muscle fibers filled with collagen fibers. The cross-sectional areas of the muscle fibers of both groups are shown in figure 4B, with significantly different values between nerve-to-nerve and nerve-to-muscle sutures ($p \lt 0.01$ ). On a macroscopic level, figures 4C–​4G present the wet weight of BBMs and the intramuscular nerves distribution (IND). The wet mass of BBMs in the nerve-to-nerve group is significantly higher than that in the nerve-to-muscle group ($p \lt 0.05$ ). Figure 4D and 4F, demonstrate the morphology of the IND in the control BBMs. Before entering a muscle, the nerve trunk emits multiple small branches, which continue to give rise to progressively smaller branches after entering the muscle. Among the BBMs in nerve-to-muscle and nerve-to-nerve groups, only nerve-to-nerve group exhibited a distribution pattern similar to the multi-level branching observed in the control BBMs (Figure 4G). In nerve-to-muscle group, the implanted nerves only emitted small branches within a small area around the joint, and no such small branches were observed to connect with the intramuscular nerve.

Fig. 4.

Physiological conditions of muscles and TMR intramuscular nerve distribution (IND). (A) The cross-section staining of targeted muscle reinnervation (TMR) biceps brachii muscles (BBMs) in nerve-to-muscle and nerve-to-nerve models. (B-C) The difference of myocytes’ size and preservation ratio of TMR BBMs in nerve-to-muscle and nerve-to-nerve models. (D and F) Control BBMs in nerve-to-muscle and nerve-to-nerve models stained with Sihler’s technique to show IND. Filled and unfilled triangles represent the original nerve and its sub-branches, respectively. (E) IND of the TMR BBM with nerve-to-muscle suture. Filled square and dotted circle represent the original nerve and its regeneration, respectively. (G) IND of the TMR BBM with nerve-to-nerve suture. Filled square, rhombus, solid circle, and unfilled square represent the transferred UN, MCN, nerve coaptation site, and original sub-branches, respectively. Scale in yellow: $400~\mu$ m, scale in black: 5 mm. (${}^{\ast } \text {)}$ there is a significant differences at ${p} \lt 0.05$ , (${}^{\ast \ast }\text {)}$ there is a significant differences at ${p} \lt 0.01$ .

SECTION IV. Discussion

In terms of TMR surgery, there are two ways for reconstructing myoelectric sites: nerve-to-muscle suture is to directly transfer a nerve onto a targeted muscle [8], while nerve-to-nerve suture is to make end-to-end anastomosis between a nerve and the original nerve of the targeted muscle [16].

To assess the function of the nerve-muscle system for TMR animals, we conducted an analysis of the characteristics of EMG signals. The RMS of EMG signals for the TMR BBMs in both nerve-to-muscle and nerve-to-nerve models did not exhibit significant improvement until 14 days post-operation; however, over time, both surgical approaches showed a progressive increase in amplitude. Regarding the SNR, the initial recording on the 7th day showed no significant difference between nerve-to-muscle and nerve-to-nerve BBMs. Yet, subsequent recordings on the 14th, 21st, and 28th days from nerve-to-nerve BBMs demonstrated significantly higher SNR compared to nerve-to-muscle BBMs. This increasing trend in nerve-to-nerve models is consistent with the findings, where the RMS of EMG signals gradually increased within four weeks following TMR surgeries [27]. A high SNR in EMG signals is crucial for myoelectric prosthesis control as it enhances signal quality and improves the accuracy of movement intention recognition [28]. Additionally, a high SNR also increases the robustness of the system, making the signal more stable and clearer, and reducing the impact of noise on signal decoding. As a result, even in different usage environments or when muscle signals are weak, the system can still reliably and accurately execute the user’s intentions, ensuring the dependability and precision of prosthesis control [29].

It is noteworthy that individual variations in EMG signals amplitude exist, as illustrated in Figure 2B and 4D, representing the control EMG signals in nerve-to-muscle and nerve-to-nerve models, respectively. To impartially evaluate the two TMR methods, EMG signals from the TMR BBMs were normalized by dividing them by the contralateral control EMG signals. After normalization, a strong correlation emerged between EMG signals amplitude and EMG signals SNR. Specifically, only the initial recording showed no significant difference, whereas subsequent recordings from nerve-to-nerve models on the $14^{\mathrm {th}}$ , $21^{\mathrm {st}}$ , and $28^{\mathrm {th}}$ days were notably higher than those from nerve-to-muscle models. In the frequency domain, concerning the CF, all four recordings from nerve-to-nerve BBMs were significantly higher than those from nerve-to-muscle BBMs, with no significant difference observed between nerve-to-nerve BBMs and control BBMs. The authors speculate that this might be related to the velocity or number of TMR muscle fibers each method can innervate. The nerve-to-nerve suture may produce a faster velocity or a larger number of muscle fibers. The faster the muscle fiber conduction velocity, the higher the frequency of the EMG signals, which represents a positive association during isometric contractions sustained at a constant force [30]. In addition, we think that the muscles in the nerve-to-muscle group are more prone to fatigue, as fatigue also results in a decrease in the frequency-domain metric of the EMG signals [31], [32].

When comparing the morphology of intramuscular nerves in different sutures, it became evident that the nerve-to-nerve suture led to a more tree-like distribution of nerves in the BBMs, similar to that observed in the control BBMs. This suggests superior nerve regeneration compared to the nerve-to-muscle suture, which resulted in a smaller area of stained nerves indicating significant nerve degeneration. A previous study have utilized a nerve-to-nerve suture, suturing the median, ulnar, and radial nerves from the brachial plexus to different segmental motor branches of the rectus abdominis muscle, successfully reconstructing three independent EMG signals sources. Both EMG signals and glycogen depletion patterns confirmed the relative independence of these reinnervated segments. Additionally, it was reported that the number and cross-sectional area of myelinated fibers in the regenerated nerves showed no significant difference compared to healthy nerves, suggesting that the nerve-to-nerve suture supports robust nerve regeneration and effective reestablishment of EMG signals sites [17]. In contrast, nerve-to-muscle suture only involved a nerve-muscle joint, creating a less favorable environment for nerve regeneration due to the natural differences between nerve tissue and muscle. A study using a nerve-to-muscle suture, where a single nerve was coapted to a muscle, demonstrated the successful reconstruction of a single EMG signals site, with muscle fiber type populations transforming to match the characteristics of the donor nerve’s original muscles [18].

A major reason is proposed to elucidate the superior outcomes of the nerve-to-nerve suture. The micro-environment in which the nerves regenerated was more conducive in the nerve-to-nerve suture. In this type of repair, both axons and myelin in the distal stump were effectively cleared by macrophages, leading to the proliferation of Schwann cells and the up-regulation of neurotrophic factors, and forming Büngner bands to facilitate axon growth [33], [34]. The study found that nerve-to-nerve BBMs had a stronger MSCF compared to nerve-to-muscle BBMs, which may be due to its richer IND and less muscle atrophy. Additionally, the TCF status of nerve-to-nerve BBMs was better than of nerve-to-muscle BBMs, characterized by shorter RD and stable properties. This suggests that nerve-to-nerve BBMs may provide consistently stable motor information and prevent muscle fatigue better than nerve-to-muscle BBMs. The combination of fatigue susceptibility and the decline in the CF of the EMG signals produced by nerve-to-muscle BBMs further demonstrates that during muscle fatigue, the conduction velocity of muscle fibers decreases, and the CF of the EMG signals shift toward the lower frequency range [31].

Furthermore, analyzing the myocytes’ size of the TMR BBMs showed that in nerve-to-nerve group, it had a significantly greater area than that of nerve-to-muscle group. Additionally, the PR of wet mass for TMR BBMs in nerve-to-nerve group was found to have no difference compared to control BBMs, but it was considerably heavier than that of in nerve-to-muscle group. Both the microcosmic muscle cell and macrograph musculature mass suggested that TMR muscles status in nerve-to-nerve group were better than that of in nerve-to-muscle group. However, both nerve-to-muscle and nerve-to-nerve sutures sacrificed the motor nerve with different forms of restoration, leading to immediate weakness of voluntary function and loss of mass in the skeletal muscle due to injury to its motor nerve over time [35], [36]. Based on the analysis above, it was observed that the recovery time for nerve-to-muscle group was longer than that of nerve-to-nerve group. This meant that the TMR muscles in nerve-to-muscle group had a longer time in the absence of full or partial reinnervation, eventually leading to worse musculature physiology statuses that were harmful to axonal regeneration [33].

Nerve-to-nerve suture could produce better motor function outcomes compared to nerve-to-muscle suture, while the latter is more commonly used in clinical TMR cases due to anatomical constraints. These constraints include insufficient nerve diameter, length, and quantity in the targeted muscle, making nerve-to-nerve approach less feasible in clinical cases, which highlights a key advantage of TMR. Such as in patients with shoulder disarticulation, where the diameter of the free nerve in the stump is much larger than that of the motor nerves of the pectoralis major or serratus anterior, leading to mismatch between donor nerve and recipient nerve [8]. To enhance nerve growth and reduce muscle atrophy, various interventions could be employed, such as electrical stimulation to slow muscle atrophy and low-intensity focused ultrasound stimulation modulates nerve regeneration, or the administration of exercise to accelerate nerve regeneration and reduce muscle atrophy [37], [38], [39]. In terms of producing additional EMG signals for amputees, regenerative peripheral nerve interface (RPNI) is also a choice [40]. However, they differ significantly in their surgical procedures: TMR involves transferring the nerve to a targeted muscle while preserving the muscle’s anatomical structure, whereas RPNI keeps the stump nerve in place and wraps it with a devascularized and denervated muscle graft. Besides, in terms of EMG signals acquisition, TMR can utilize both surface electrodes and intramuscular electrodes, whereas RPNI requires the use of intramuscular electrodes due to the deep location and small size of the muscles involved.

In addition to enhancing intuitive control of myoelectric prostheses for amputees, TMR and RPNI have also been shown to alleviate neuropathic pain in residual limbs after amputation [41], [42]. A randomized clinical trial found that one year postoperatively, patients who underwent TMR experienced a more than 50% reduction in neuropathic pain scores compared to the control group [43]. Similarly, another randomized clinical study reported that 85% of patients who underwent RPNI experienced pain relief six months after surgery, with some achieving complete pain resolution [44]. TMR, RPNI and end-to-side neurorrhaphy reduce neuropathic pain by redirecting severed nerve bundles to reinnervate muscles, preventing them from forming neuromas and thereby alleviating pain [7], [14], [45], [46].

However, the study also had some limitations. The initial plan was to record EMG signals until the 90th day post-operation, while the implanted electrodes in animals tend to fail within approximately 1–2 months for various reasons. Possible causes include detachment at the skull interface, electrode wire breakage due to pulling, or encapsulation by the immune system, which obstructs signal transmission, among others. This prevented the researchers from obtaining more EMG signals. As a result, only one month of EMG signals were available. Usually the recovery of peripheral nerves in rats typically takes around three months to reach its maximum. Therefore, we believe that with the extension of rehabilitation time, the EMG signals of both methods may improve and be enhanced. In the future, we will improve long-term implantation methods, extend the implantation time, and collect EMG signals over a longer period, with the aim of further comparing the final effects of these two methods on the characteristics of EMG signals. Besides, the long-term stable communication interface technology has been proposed, leveraging osseointegration to implant myoelectric electrodes directly into targeted muscles. The electrode wires transmit EMG signals through the bone-integrated implant. Clinical studies conducted so far have achieved continuous EMG signals monitoring for up to 48 weeks, with a SNR of 20 dB and a motion intent recognition accuracy of 97% by the last follow-up [47]. Although this study analyzed several characteristics of EMG signals, the lack of behavioral training in the experimental animals limited the application of these features to the research on motion intention recognition. Future studies should incorporate systematic behavioral training to further validate the effectiveness and potential applications of these features in motion intention recognition. Additionally, the study only focused on motor nerve regeneration and did not study sensation nerve regeneration, such as pain administration. Future research can focus on comparing nerve-to-muscle and nerve-to-nerve sutures based on sensation nerve regeneration as well.

SECTION V. Conclusion

EMG signals amplitudes gradually increased over time in both groups, with a significant difference observed—nerve-to-nerve suture resulted in higher amplitudes than nerve-to-muscle suture. Furthermore, the SNR and CF of EMG signals were notably higher in the nerve-to-nerve models compared to the nerve-to-muscle models. These findings suggest promising potential for clinical applications.