Changes in tibialis anterior architecture affect the amplitude of surface electromyograms

Electrical stimulation was applied to the fibular nerve of participants’ right leg, while they stood upright and with their foot secured to a force meter (PY6, Bertec, USA). Monophasic current pulses (200us; Digitimer, UK) were delivered to the fibular nerve with pre-gelled, closely spaced electrodes (Fig. 1a; electrode size: 10 × 10 mm; 10 mm inter-electrode distance). Stimulation electrodes were positioned at the skin region closest to the fibular nerve, where the least injected current led to clearly observable, TA twitches.

Two sets of ten current pulses (2 pps) each were considered to elicit TA M waves, at 50% and 100% of the current intensity (range: 30–55 mA) over which no further increases in force could be observed. We decided to define the maximal current intensity based on force twitches rather than on M-wave amplitude because establishing the maximal M-wave amplitude from surface EMGs seems unviable for in-depth, pennate muscles [15]. Electrical stimulation was applied for three different, ankle joint angles. While standing with their right foot secured to the floor using straps, subjects were specifically asked to position their left foot (Fig. 1b): i) as much as possible backward (plantar flexion position; PF); ii) alongside the right foot (neural position; NE); iii) as much as possible forward (dorsiflexion; DF). The experimenter assisted subjects in changing position without losing balance and therefore without moving their right foot. A total of 60 stimulation pulses were applied per subject (10 pulses × 2 stimulation intensities × 3 ft positions). Stimulation order was randomised and 2 min rest was provided between stimulation sets.

Monopolar EMGs were detected with a grid of 64 surface electrodes (16 × 4 electrodes; 8 mm diameter; 10 mm inter-electrode distance). Electrodes were secured to the skin with bi-adhesive foam and the electrode-skin contact was ensured by filling the foam cavities with conductive paste. The grid was positioned at a skin region covering as much as possible the TA superficial aponeurosis (Fig. 1a); the most medial column of electrodes was located 10 mm laterally to the tibial crest and the most proximal row was located 10 mm distally to the head of the fibula. Skin was cleaned with abrasive paste prior to placing the electrodes. EMGs were detected with a multi-channel fixed gain amplifier (W-EMG amplifier, LISiN-Politecnico di Toino, Italy; [16]). Trigger pulses indicating the onset of stimulation pulses were digitised synchronously with EMGs at 2441 Samples/s (24 bit A/D converter).

Ultrasound (US) images were recorded (Echoblaster 128, Telemed, LT) using a 4 cm linear array transducer (10 MHz) in both transverse and longitudinal planes [17]. The transducer was positioned tangentially to the skin, centred along the leg circumference defined by the fifth row of electrodes (Fig. 1a). Care was taken to ensure minimal pressure and thus to avoid distortion of image features related to the muscle tissue [17]. When recording transverse images, a series of contiguous images was sampled to provide a view of the entire muscle width. Longitudinal images were taken from the skin region between the second and third columns of electrodes, while ensuring fascicles could be clearly visualized from the deep to the superficial aponeurosis [18]. Images were collected for each of the three ankle positions, while subjects stood as shown in Fig. 1b. Subjects were instructed to relax their muscle as much as possible during both ultrasound and M-waves’ recording, otherwise changes in TA architecture resulting from changes in the degree of muscle contraction [19] would likely affect M waves. We ensured subjects held their muscle relaxed during electrical stimulation through visual inspection of surface EMGs; no single action potentials were observed during stimulation.

A cluster consisting of four reflective markers attached to the US probe was considered for calibration; 3D position of the US image plane in the cluster reference frame was calibrated based on previous reports [20, 21]. Eight reflective markers were attached on anatomical landmarks of the individuals’ right leg and foot (tibial tuberosity, head of the fibula, lateral malleolus, medial malleolus, first-, second-, fifth metatarsal and calcaneus) as indicated by Cappozzo and colleagues [22]. To ensure that the static ankle position was maintained during data acquisition and to track the cluster position of the US probe, kinematic data were collected at 100 Hz and offline synchronised via an external trigger.

M waves detected with electrode grid: EMGs were first band-pass filtered with a fourth order, Butterworth filter (15–350 Hz cut-off frequencies). M waves were then obtained by averaging EMG samples over 30 ms epochs, with epochs starting from the onset of each trigger pulse; EMGs were averaged across pulses, separately for each electrode, stimulation intensity and ankle position (Fig. 2). The number of electrodes detecting relatively large M waves (termed active channels) and the region where these electrodes were located (the transverse and longitudinal coordinates of active channels’ centroid) were computed from both the M-waves peak-to-peak and root mean square amplitude. Active channels were identified by applying an automated technique to segment EMG images [23] and centroid coordinates were calculated as the weighted average of active channels across rows and columns (cf. grey and crossed circles in Fig. 2). Only electrodes located over the TA superficial aponeurosis were considered for analysis, given the amplitude distribution of EMGs detected in such region reflects the global degree of muscle activation [24].

M waves detected with bipolar electrodes: the effect of ankle position was tested on the peak-to-peak amplitude of M waves detected by the conventional, bipolar montage. Bipolar signals were simulated as the sample-by-sample difference between monopolar M waves averaged over two sets of four electrodes (Fig. 1a), centred in the location currently recommended for the positioning of bipolar electrodes on TA [25].

TA width and pennation: changes in TA width and pennation angle were estimated from US images. US images, acquired contiguously and transversally from TA, were projected in the 3D laboratory reference frame using the cluster probe pose and the previously performed calibration [20]; then, the images were superimposed and concatenated to provide a view of the entire muscle transverse section. Width was computed as the distance between the medial and lateral TA borders. Pennation angle was calculated from parasagittal images [26], as the acute angle between two intersecting lines representing fascicle orientation (line of the clearest fascicle) and deep aponeurosis [26].

Joint angles: Ankle angles were calculated from kinematic data, considering a 2-segment model of the right lower limb [27, 28]. After ensuring that static position was maintained during each test (standard deviation inferior to 1°), mean ankle angles were calculated for each subject and each position.

Ankle DF corresponded to approximately 30° of flexion and PF to 40° of extension. Detailed pose data are shown in Table 1. When considering M waves, peak-to-peak and root mean square amplitude values provided statistically equivalent results. For convenience, we report results for the peak-to-peak amplitude.

The effect of foot position on TA architecture and M waves is well appreciated by inspecting results from a single, representative participant. As shown in Fig. 2, the TA pennation angle increased from 5.5 to 13.0 deg. when this participant moved his right foot from PF to DF position (Fig. 2). Clearer M waves with greater amplitude were elicited with the foot in DF and NE rather than in PF position, regardless of where they were detected in the grid. Regional differences in M-wave amplitude were observed however, in particular for the DF position; in such position the similarity between M waves detected by different channels was less evident (cf. M waves and images in Fig. 2).

When considering all participants, ankle position affected significantly M-waves’ peak-to-peak amplitude though not its distribution on the skin. M-waves’ amplitude was significantly smaller in PF than in NE and DF positions, both for 50% and 100% of the maximal stimulation intensity (Fig. 3a; ANOVA; P < 0.001). The size (i.e., number of active channels) and centre of M-waves’ amplitude distribution were however not affected by ankle position (P > 0.31), even though relatively larger peak-to-peak values tended to be detected over a larger and distal region in PF (Fig. 3b–d). Stimulation intensity, on the other hand, affected both the mean peak-to-peak amplitude (Fig. 3a) and the number of active channels (Fig. 3b; P < 0.038 in both cases), with smaller and more spatially diffused peak-to-peak values being obtained in PF position.

As for monopolar signals, when moving the foot from dorsi to plantar flexion, the bipolar M waves became smaller and wider (cf. waveforms in Figs. 2 and 4a). When compared to DF and NE, the PF position provided M waves with significantly smaller peak-to-peak amplitude for the two stimulation intensities considered (Fig. 4b; P < 0.001). Moreover, for all ankle positions, peak-to-peak values of bipolar waves almost doubled from 50% to 100% stimulation intensity (P < 0.001).

Muscle architecture changed with foot position. As shown in Table 2, pennation angle and muscle width increased significantly with the ankle dorsiflexion.

The amplitude of M waves was significantly associated with both TA pennation angle and width. Greater pennation angle and width resulted in M waves with greater amplitude (Fig. 5), both during 100% (Pearson R > 0.36; P < 0.04; N = 30; 10 subjects × 3 ankle angles) and 50% stimulation intensities (Pearson R > 0.42; P < 0.025; N = 30). Similar, significant results were observed for bipolar M waves (Person R > 0.35 and P < 0.035 in all cases).

Three considerations are necessary before interpreting current results. First, the effect of changes in muscle architecture on the amplitude of surface EMG was assessed during electrically elicited rather than during voluntary contractions. In the latter case, the sensitisation of joint, muscle and tendon receptors, resulting from changes in foot position, could lead to changes in muscle activity [29]. During voluntary contractions, it would thus not be possible to dissociate the effect of architectural changes from that of physiological mechanisms on the amplitude of TA M waves. Second, changes in TA architecture were induced by changing foot positioning while keeping the knee joint angle constant (Fig. 1a). We decided to keep the same leg-thigh orientation for the different ankle joint angles to ensure minimal, potentially negligible, changes in the relative position between the fibular nerve and stimulation electrodes. Even if the fibular nerve had moved in relation to the stimulation electrodes when moving the foot from DF to PF, these relative movements would be likely random and therefore would not explain the consistent changes in M-wave amplitude with foot positions across participants (Figs. 3, 4). Finally, TA M waves appear a few milliseconds after current pulses are delivered to the fibular nerve and last roughly 20 ms (cf. Fig. 2; [30]). Considering the force twitches develop within somewhat longer periods, the M waves obtained here are therefore unlikely affected by TA architectural changes associated with the force twitches. Collectively, these arguments suggest the results reported here are presumably not due to factors other than the change in TA architecture.

The relationship between architectural changes and EMG amplitude is likely muscle dependent. For muscles with fibres oriented parallel to the skin, for example, it is well established that shifts in the innervation zone, resulting from changes in joint angle, affect the amplitude of surface EMGs; EMG amplitude is greatly underestimated when detected nearby the innervation zone [9, 10]. For in-depth pennate muscles, such as gastrocnemius and TA, the effect of changes in muscle architecture on the surface EMG is less well documented. Comparing current and previous results is therefore currently not possible. Recent, preliminary results, obtained with low intensity nerve stimulation, suggest however the amplitude distribution of surface EMGs changes with the gastrocnemius pennation angle [13]. For greater degrees of pennation, the projection of gastrocnemius fibres on the skin decreases, leading to the detection of greater EMGs at more localised and proximal gastrocnemius regions (cf. Fig. 4 in Mesin et al. [13]). Current results on TA partly corroborate these previous findings; we observed TA pennation affected M-wave amplitude though not its distribution (Figs. 2, 3, 4). Likely because we excited a somewhat great proportion of muscle fibres, changes in the distribution of M-wave amplitude with variation in foot position did not reach statistical significance (Fig. 3b–d). Additionally, differently from gastrocnemius, TA fibres are roughly symmetrically angled with respect to the muscle intermediate aponeurosis (Fig. 2; [31]). In virtue of the depth of the TA intermediate aponeurosis, action potentials of fibres running from the TA deep to intermediate aponeurosis would be expected to affect surface EMGs equally across electrodes (far-field potentials; [32]). Although we value the relevance of identifying the mechanisms underpinning the relationship between TA architecture and EMGs, here we were focused on understanding whether rather than how EMGs are affected by changes in TA architecture. Notwithstanding the potential causes, here we show that for a constant TA excitation surface EMGs with different amplitudes are detected for different foot positions.

Regardless of whether taken locally or not, EMG amplitude was affected by TA architectural changes. Given grids of electrodes sample from a somewhat large muscle region, they overtly provide surface EMGs less sensitive to spurious factors than the conventional, bipolar derivation [9, 15]. Here we show however the effect of TA architectural changes maybe not compensated for by sampling EMGs everywhere over the TA superficial aponeurosis; M-wave amplitude increased with the foot dorsiflexion for both recording modalities (Figs. 3a, 4b and 5). Therefore, surface EMGs should be interpreted carefully when collected from TA both with bipolar electrodes and with grids of electrodes.

Assumptions are often made whenever EMGs are used to advance considerations on the muscle function. Even though it is usually assumed the surface EMGs sample exclusively from the whole volume of target muscles, it has been shown the surface EMGs may sample from muscles other than that of interest (i.e. crosstalk; [30]) or, conversely, from a small, unrepresentative muscle region [15, 24]. In both cases, drawing inferences on the whole muscle level from single, localised EMG recordings maybe misleading. Of specific, practical interest here is the assumption that variations in the amplitude of surface EMGs indicate variations in the neural drive to the muscle; i.e., in the recruitment and firing rate of motor units. More specifically, increased and decreased amplitude of surface EMGs collected from a given muscle during a given condition are respectively conceived as augmented and diminished degree of muscle activity [4, 5]. The present study shows (Figs. 3 and 4) that this hypothesis is not always valid. Here we show that from 13% to 25% of the changes in the amplitude of TA M waves may be explained by changes in TA architecture (Fig. 5); even in the presence of a constant degree of excitation, M-wave amplitude was roughly 25% smaller in PF than in DF and NE positions (Fig. 3). When considering the conventional, bipolar montage, differences in M-wave amplitude with foot position increased to ~50% (Fig. 4). These results raise the question as to which extent variations in EMG amplitude observed in circumstances imposing TA architectural changes maybe attributable to physiological sources. In conclusion, while the possibility of compensating for these architectural effects on the EMGs urges further investigation, variations in EMG amplitude during dynamic conditions should be interpreted cautiously.

The generalisation of results to different muscles and circumstances is a key limitation of this study. Our decision to assess the TA muscle was motivated by its key role in human motion analysis [5, 26]. Although the variations in EMG amplitude with changes in foot position agree with previous observations for the gastrocnemius muscle [13], we refrain from stating current results generalise to in-depth pennate muscles other than TA. Anatomical differences between muscles (e.g. different degrees of pennation, or pennation in oblique planes) could lead to relationships between EMG and muscle architecture different from that reported here (Figs. 2, 3, 4, 5). Finally, it should be noted we may have underestimated the effect of TA architecture on M-wave amplitude. Considering the TA pennation angle increases with the degree of TA activity [14, 19], it is possible that M-waves’ amplitude would have increased to greater extents had we applied current pulses while subjects were sustaining a certain degree of TA activation. The caveat though is isolating the effect of TA architecture on the amplitude of M waves during voluntary contraction, as discussed above. While the generalisation of current results remains the subject of future investigations, it seems however likely that, for different muscles and conditions, changes in EMG amplitude result from structural in addition to physiological changes.