Nature of the coupling between neural drive and force-generating capacity in the human quadriceps muscle

Abstract

The force produced by a muscle depends on both the neural drive it receives and several biomechanical factors. When multiple muscles act on a single joint, the nature of the relationship between the neural drive and force-generating capacity of the synergistic muscles is largely unknown. This study aimed to determine the relationship between the ratio of neural drive and the ratio of muscle force-generating capacity between two synergist muscles (vastus lateralis (VL) and vastus medialis (VM)) in humans. Twenty-one participants performed isometric knee extensions at 20 and 50% of maximal voluntary contractions (MVC). Myoelectric activity (surface electromyography (EMG)) provided an index of neural drive. Physiological cross-sectional area (PCSA) was estimated from measurements of muscle volume (magnetic resonance imaging) and muscle fascicle length (three-dimensional ultrasound imaging) to represent the muscles' force-generating capacities. Neither PCSA nor neural drive was balanced between VL and VM. There was a large (r = 0.68) and moderate (r = 0.43) correlation between the ratio of VL/VM EMG amplitude and the ratio of VL/VM PCSA at 20 and 50% of MVC, respectively. This study provides evidence that neural drive is biased by muscle force-generating capacity, the greater the force-generating capacity of VL compared with VM, the stronger bias of drive to the VL.

1. Introduction

The torque produced by a muscle depends on both the neural drive it receives and several biomechanical factors such as its physiological cross-sectional area (PCSA), force–length and force–velocity relationships, specific tension, and moment arm. Human experiments have provided some evidence that the mechanical advantage of a muscle (related to its moment arm) and the neural drive it receives during a voluntary motor task is coupled [1–3]. For example, during a task where the mechanical advantage of the first dorsal interosseous muscle is increased by increasing its moment arm through altered thumb posture, there is an adaptive increase in the neural drive to that muscle [3]. This adaptive change in drive to match mechanical efficacy is not universal; when the force-generating capacity of a muscle is acutely decreased by muscle damage, activation of all synergist muscles (including the injured muscle) increases rather than recruitment of only the non-injured muscles [4]. This strategy appears suboptimal and may be explained by the limited potential for a short-term redistribution of neural drive between some synergist muscles [5,6]. However, it is possible that the coupling between drive and muscle force-generating capacity adapts over time, with practice and repetition. Understanding the nature of the coupling between neural drive and muscle force-generating capacity is a critical step toward a deeper understanding of movement control in humans.

Although many biomechanical models consider that the neural drive is equally shared between synergist muscles during submaximal tasks [7,8], relatively large differences in motor drive (determined from electromyography (EMG) amplitude) between synergist muscles and between participants has been demonstrated. For example, during walking greater EMG amplitude is observed in the medial than the lateral gastrocnemius in ≈50% of participants, whereas the other half has similar EMG amplitude of both muscles [9]. Similarly, during a wide range of tasks, some [10,11], but not all [12], studies report greater EMG amplitude of the lateral (vastus lateralis, VL) than the medial (vastus medialis, VM) head of the quadriceps with a large difference between individuals (e.g. ratio of VL/VM activation ranges from ≈0.8 to 3 [10]).

Similarly, muscle PCSA, which is a key determinant of force-generating capacity [13,14], is variable between individuals. For example, a cadaveric study reported variation in the ratio of VL/VM PCSA from 0.9 to 2.2 [15]. The nature of the coupling between neural drive and muscle force-generating capacity is unknown. We have three hypotheses. First, neural activation is adjusted to balance forces between synergist muscles of differing force-generating capacities, in which case the muscle with the lower force-generating capacity would receive greater neural drive. Second, neural activation is targeted to reduce the overall neural cost, in which case the muscle with the higher force-generating capacity would receive greater neural drive. Third, there is no relationship between neural drive and force-generating capacity. In the latter two cases, the force produced by the two synergist muscles would be imbalanced. Imbalance such as this has been proposed to underpin the development of some musculoskeletal conditions. For example, an imbalance of force generation between VL and VM has been speculated to contribute to the development and/or persistence of patellofemoral pain [16].

The primary aim of this study was to determine the relationship between the ratio of neural drive measured during isometric tasks and the ratio of muscle torque-generating capacity between two mono-articular heads of the quadriceps muscle group in humans, i.e. VL and VM. Muscle torque-generating capacity was determined from the muscle's PCSA. As the moment arm relative to the midpoint of the tibiofemoral flexion/extension axis is not different for these muscles [17,18], it was not considered in this study. For the relationship between neural drive and muscle PCSA to be meaningful, we first needed to demonstrate that the distribution of neural drive between the heads of the quadriceps is repeatable between days.

2. Material and methods

(a). Participants

Twenty-two healthy volunteers (mean ± s.d., age: 27 ± 7 years, weight: 69 ± 12 kg, height: 175 ± 7 cm; 11 females) participated in this study. Participants had no history of knee pain that had limited function or required them to seek intervention from a healthcare professional.

(b). Assessment of neural drive

(i). Experimental set-up

To test the between-day reliability of measures of VL and VM activation, participants attended two identical testing sessions interspaced by 1–2 days. Participants sat on a plinth with their back and upper legs supported. The torso was reclined by 10° from upright and the knee (dominant leg, i.e. left leg for four participants) flexed to 80° from the horizontal. This angle was chosen as it represents the optimal vastii muscle length, i.e. the knee angle at which the maximal knee extension torque can be generated [19]. In addition, maximal voluntary activation is the greatest at this angle [20]. A standard 5 cm wide support strap was placed around the pelvis to minimize changes in body position throughout the experimental task and participants crossed their arms over their chest.

(ii). Surface electromyography

Myoelectrical activity was recorded from the test (dominant) leg with surface EMG electrodes placed over the VM and VL. For each muscle, a pair of self-adhesive Ag/AgCl electrodes (Blue sensor N-00-S, Ambu, Denmark) was attached to the skin with an inter-electrode distance of 20 mm (centre-to-centre) at the site recommended by SENIAM [21]. For both VL and VM, B-mode ultrasound (Aixplorer, Supersonic Imagine, France) was used to facilitate the placement of the electrodes longitudinally with respect to the muscle fascicle alignment. The electrode locations were marked on the skin using a waterproof marker to guide repeated placement between days. Prior to electrode application, the skin was cleaned with abrasive gel (Nuprep, D.O. Weaver & Co, USA) and alcohol. The ground electrode (half of a Universal Electrosurgical Pad, 3M Health Care, USA) was placed on the skin over the tibia of the test leg. EMG data were amplified 1000 times, band-pass filtered between 10 Hz and 500 Hz (Neurolog, Digitimer, UK) and sampled at 2 000 samples s⁻¹ using a Power1401 Data Acquisition System with Spike2 software (Cambridge Electronic Design, UK).

(iii). Mechanical data

Force was measured with a six-axis force sensor (Sensix, Poitiers, France). Signals were amplified, sampled at 2 000 samples s⁻¹, and low-pass filtered at 20 Hz (Power1401 Data Acquisition System, Cambridge Electronic Design, UK).

(iv). Voluntary activation level

In order to estimate the maximal voluntary activation level, a doublet electrical stimulus (inter stimulus interval—10 ms) of 200 µs duration and 400 V amplitude was delivered by a Digitimer DS7AH constant current stimulator (Digitimer, UK) through large surface electrodes placed over the femoral triangle (one-fifth of a Universal Electrosurgical Pad) and on the gluteal fold (half of a Universal Electrosurgical Pad). The resting twitch of maximal amplitude was determined by applying stimuli of increasing intensity in steps of 10 mA, until knee extension force plateaued despite an increase in current intensity. To ensure a maximal response throughout testing, supramaximal stimulus intensity was used, corresponding to 120% of the intensity that evoked a maximal resting twitch response (mean: 271 ± 84 mA, range: 132–444 mA).

(v). Experimental tasks

After a standardized warm-up and prior to commencement of the experimental trial, six maximal isometric voluntary knee extensions were performed against a rigid resistance. Each contraction was maintained for 3 s and separated by 120 s rest. Maximum knee extension torque was considered the highest performance (maximum voluntary contraction (MVC) torque). In order to verify that participants produced true maximal activation and thus confirm that both VL and VM were maximally activated, we used the twitch interpolation technique during the last three maximal knee extensions to quantify the voluntary activation level. A supramaximal doublet stimulus was delivered during the plateau of MVC when at maximal amplitude, and within 5 s in the subsequent rest period to elicit superimposed and rest twitches, respectively. Then, the experimental task involved matching submaximal target force set at 20 and 50% of MVC during short (≈10–15 s) isometric knee extensions with 30 s rest between each repetition. Contractions were repeated three times at each intensity. The order of the contractions was randomized.

(vi). Force and electromyography data analysis

All force and EMG data were processed using Matlab (The Mathworks, Nathick, USA). MVC force was calculated from the first three maximal knee extensions as the maximal force measured over a 300 ms time window. The percentage of voluntary activation was measured during the last three maximal knee extensions according to the equation given by Todd et al. [22]:

2.1

The ‘superimposed twitch’ amplitude was defined as the difference between the peak torque induced by the stimulation and the maximal voluntary force averaged over 300 ms before the stimulation.

To determine the maximal EMG amplitude achieved during the first three maximal knee extension tasks, the root mean square (RMS) of the EMG signal was calculated over a time window of 300 ms and the maximal value was considered as the maximal activation level. During the submaximal isometric knee extensions, the RMS EMG amplitude was calculated over 5 s at the middle of the force plateau. These values were normalized to the maximal RMS EMG values determined as described above. The ratio of neural drive between VL and VM was calculated from these normalized values as:

2.2

(c). Assessment of muscle physiological cross-sectional area

(i). Magnetic resonance imaging

Volumetric acquisitions of the thighs were performed using a 3 T magnetic resonance imaging (MRI) scanner (Magnetom Trio, Siemens Healthcare, Germany) with the participants in supine position, lying with their hip and knee fully extended. The three-dimensional (3D) Vibe sequence was chosen to enhance the separation between muscles, thus improving the accuracy of the segmentation (repetition time: 20 ms, echo time: 1.57 ms, field of view: 300 × 400 mm², voxel size: 0.78 × 0.78 × 5 mm, flip angle: 25°). Slice thickness was 5 mm without an inter-slice gap. Three volumes were acquired (acquisition time: 5:43 min per volume) to image the anatomical structures from iliac crests to the articular surface of the lower border of the patella. All data were transferred in DICOM format.

MR images were analysed using 3D image analysis software (Mimics, Materialise, Belgium). Both VL and VM were segmented manually and independently by two operators (C.G., D.B.). Each slice was analysed from the distal slice, where the VM could first be visualized to the most proximal slice at the level of VL insertion (figure 1). As reported previously from cadaveric studies [23], vastus intermedius and VL were fused in some slices of the more proximal regions (10 ± 8% of the slices). As proposed by Barnouin et al. [24], we maintained a ‘bean’ shape for VL and consistently used the visible landmarks on the preceding and subsequent images to assist the segmentation between muscles. The volume of each muscle was calculated from 3D reconstruction (option = optimal; Mimics).

Figure 1.

Individual example of muscle segmentation and muscle volume reconstruction. First, vastus medialis (darker grey in printed version / red in online version) and vastus lateralis (lighter grey in printed version/ orange in online version) muscles were manually segmented from each axial image (note that only four images from a total of 68 used to generate this reconstruction are depicted on the right panels). Then, muscle volume was reconstructed using 3D image analysis software (left panel). (Online version in colour.)

(ii). Fascicle length

Because of its architecture, classical two-dimensional (2D) ultrasound techniques cannot be used to reliably estimate VM fascicle length [25,26]. Therefore, we used freehand 3D ultrasound scans to measure fascicle length of both this muscle and VL. 3D ultrasound involves combining multiple 2D ultrasound images of the muscle with 3D motion data of the transducer's orientation and position to reconstruct the muscle in 3D [27,28]. 2D B-mode ultrasound images were acquired by manually moving the ultrasound transducer along the approximate midline of each muscle belly in a transverse orientation and were subsequently transformed into the global coordinate system using Stradwin software (v.5.0; Mechanical Engineering, Cambridge University, Cambridge, UK) (figure 2). B-mode ultrasound images were recorded at 10–15 Hz using a 60 mm linear transducer (L-14–5 W/60 Linear, Ultrasonix) with a central frequency of 10 MHz and maximal depth of 70 mm. Position and orientation of the transducer were recorded by tracking a cluster of four markers rigidly attached to the ultrasound transducer using a four-camera optical motion analysis system (Optitrack, NaturalPoint, USA). Prior to scanning, the relationship between the image coordinate system and the cluster coordinate system was determined using a single-wall phantom calibration protocol available in the software package and used in previous studies [27,28].

Figure 2.

Individual example of reconstructed ultrasound images from freehand 3D ultrasound for the vastus lateralis (VL) (a) and the vastus medialis (VM) (b) muscles. The best sagittal plane re-slice images were determined visually as the images that displayed the clearest continuous muscle fascicles. Generally, one or two fascicles per re-slice image were analysed. (Online version in colour.)

Participants were seated on a plinth. Their torso was reclined by 10° from upright and their knee flexed 85–90° from the horizontal. This position closely matched the position used for neural drive measurements. A 2 cm thick acoustic stand-off pad (Aquaflex, Parker Laboratories, Fairfield, NJ, USA) was positioned between the transducer and the leg to enhance visualization of the muscle belly. The transducer was translated in the distal–proximal direction along the leg, covering the midline/thickest portion of each muscle from their distal insertion to approximately 75% of femur length. The total scan time was ≈15 s and the average distance between frames was ≈1 mm. This was repeated two to four times per muscle, depending on the quality of the reconstructed sagittal plane image re-slices.

For each acquisition, the best plane re-slice images were determined visually as the images that displayed the clearest continuous muscle fascicles, which was presumed to correspond to the plane of muscle fascicles (figure 2). Because most of the fascicles exhibited a small curvature, we used a segmented line (with a spline fit) to model the fascicle and calculate its length (ImageJ V1.48, National Institute for Health, USA). The fascicle length measurements were performed by two operators (F.H. and C.G.). For VL, three to four fascicles were measured at distal and proximal locations (total of six to eight fascicles per participant). For VM, four fascicles were measured superficially (top one-third of the muscle) and two fascicles were measured deep within the muscle (about two-thirds of maximum muscle thickness). Then, values were averaged across measurements within a muscle to get a representative fascicle length measurement per muscle. The greater number of measurements taken from the superficial compared with the deep part of VM intentionally biased the average fascicle length to the more superficial values. This is because we consistently observed that the deeper fascicles are less representative of the whole muscle. It is important to note that the analysis was also performed giving the same weight to deep and superficial compartments (data not presented), and the main outcomes of this study remained unchanged. Note that 3D ultrasound data for VL of one participant exhibited some artefacts and, therefore, we used 2D ultrasound to measure the fascicle length, using methods previously used on VL [29].

(iii). Physiological cross-sectional area

PCSA was calculated from muscle volume (cm³) and fascicle length (cm) as follows [13]:

2.3

Note that the knee angle used for 3D ultrasound data collection was close to the optimal knee angle (see above) and, therefore, the measured fascicle length was considered as the optimal fascicle length. Then, we calculated the ratio of PCSA between VL and VM as:

2.4

(d). Statistics

Note that after the first session, one participant developed knee pain and was not able to perform the second session. This participant was excluded from the experiment and data are, therefore, reported for 21 participants. Further, significant co-contraction of the hamstring muscles was induced by the superimposed twitch in one participant. The maximal voluntary activation data are reported for 20 participants.

Statistical analyses were performed in Statistica (Statsoft, USA). Distributions consistently passed the Shapiro–Wilk normality test and all data are reported as mean ± s.d. The level of significance was set at p < 0.05. The MVC torque determined from the first three maximal knee extensions (without twitch interpolation) was compared with the MVC torque determined from the next three maximal knee extensions (with twitch interpolation) using a paired t-test. The potential for difference in normalized RMS EMG and PCSA between muscles was also considered using a paired t-test. To determine the robustness of the EMG measures between the two sessions and the inter-operator reliability for MRI and 3D US data, the intra-class correlation coefficient (ICC), the standard error of measurement (SEM), and the coefficient of variation (CV) were calculated. Finally, the relationship between the VL/VM EMG ratio and VL/VM PCSA ratio was tested using Pearson's correlation coefficient. As proposed by Cohen [30], we report a correlation of 0.5 as large, 0.3 as moderate, and 0.1 as small.

3. Results

(a). Neural drive

The MVC torque measured during the first three maximal isometric knee extensions performed without twitch interpolation (236 ± 83 Nm (range: 118–418 Nm)) was not significantly different to that measured just before the twitch interpolation during the subsequent three maximal contractions (233 ± 87 Nm (range: 111–420 Nm); p = 0.084). For both sessions, all participants (except one participant for session 2) reached a voluntary activation greater than 90% (mean = 98.8 ± 2.8% (range: 90.7–100%) and 97.6 ± 3.6% (range: 87.3–100%) for sessions 1 and 2, respectively). The maximal EMG amplitude measured over the first three maximal contractions was, therefore, considered to represent the maximal neural drive to both VL and VM.

For the isometric knee extensions performed at both 20 and 50% of MVC, the inter-day reliability of the normalized VL and VM RMS EMG amplitude was good to excellent (ICC > 0.67 and SEM < 4.2% of MVC; table 1). Notably, reliability was less for VM than for VL, but remained good (table 1). The ratio of VL/VM EMG amplitude demonstrated good to excellent reliability between days (ICC > 0.74), indicating that the participants adopted a robust coordination strategy between these synergist muscles. Considering this overall good to excellent reliability, normalized EMG data were averaged between the two days for further analysis.

Table 1.

Neural drive to the vastus lateralis (VL) and vastus medialis (VM) muscles during the isometric knee extension tasks. ICC, intra-class coefficient correlation; SEM, standard error of measurement; MVC, maximum voluntary contraction.

	isometric knee extension (20% MVC)			isometric knee extension (50% MVC)
	VL EMG (% max)	VM EMG (% max)	VL/VM ratio	VL EMG (% max)	VM EMG (% max)	VL/VM ratio
day 1	15.9 ± 6.0	13.8 ± 5.9	53.8 ± 11.4	40.0 ± 9.7	37.5 ± 7.9	51.5 ± 5.8
day 2	14.2 ± 5.1	12.9 ± 6.3	53.1 ± 11.0	38.0 ± 9.3	36.6 ± 9.1	50.9 ± 6.4
ICC	0.83	0.67	0.86	0.94	0.68	0.74
SEM	2.6	3.5	4.2	2.2	4.9	3.1

The mean ratio of VL/VM EMG amplitude was 53.5 ± 11.1% (range: 33.6–74.7%) and 51.2 ± 6.1% (range: 42.2–64.7%; figure 3) for the isometric knee extension performed at 20 and 50% of MVC, respectively. Notably, there was large variability between individuals (figure 3b,c) with an almost equal number of participants demonstrating greater VL RMS EMG and those demonstrating greater VM RMS EMG. Consistent with this, the normalized RMS EMG amplitude was not different between VL and VM at either 20% (p = 0.24) or 50% of MVC (p = 0.33).

Figure 3.

Relationship between the ratio of neural drive and the ratio of muscle force-generating capacity between the vastus lateralis (VL) and vastus medialis (VM) muscles. (a) Group distribution of the ratio of physiological cross-sectional area (PCSA) between VL and VM. (b,c) Group distribution of the ratio of neural drive between VL and VM measured during submaximal isometric knee extensions performed at 20% (b) and 50% (c) of maximum voluntary contraction (MVC). (d,e) The correlation between the ratio of neural drive and the ratio of PCSA. The strong (d) and moderate (e) correlations indicate that the greater the force-generating capacity of VL compared with VM, the stronger bias of drive to the VL. (Online version in colour.)

(b). Physiological cross-sectional area

The inter-operator reliability of MRI manual segmentation was excellent for both VL (ICC = 0.99, SEM = 12.3 cm³, and CV = 2.0%) and VM (ICC = 0.99, SEM = 6.2 cm³, and CV = 1.7%). The inter-operator reliability of fascicle length measurements was also excellent for both VL (ICC = 0.95, SEM = 0.29 cm, CV = 2.5%) and VM (ICC = 0.98, SEM = 0.25 cm, CV = 2.0%). Consequently, volume and fascicle length data were averaged between the two operators for further analysis.

Muscle volume was systematically larger for VL than VM (mean difference: 35.2 ± 18.6%, p < 0.0001; table 2). As VM muscle fascicles were longer than those of VL (7.8 ± 9.8%, p = 0.007, table 2), VL PCSA was systematically larger than VM PCSA (45.2 ± 20.3%, p < 0.0001; table 2). The ratio of PCSA between VL and VM was consistently greater than 50%, i.e. 59.0 ± 3.4%. There was a large variability between participants with values ranging from 52.9 to 64.3% (figure 3).

Table 2.

Morphological and architectural data for the vastus lateralis (VL) and vastus medialis (VM) muscles. For comparison with previously published studies, data are presented separated by gender. fl, optimal fascicle length; PCSA, physiological cross-sectional area.

	vastus lateralis			vastus medialis
	males	females	mean	males	females	mean
volume (cm3)	702.0 ± 127.3	436.7 ± 58.8	563.0 ± 165.7	536.6 ± 70.0	319.7 ± 60.8	422.9 ± 128.0
fl (cm)	12.7 ± 1.2	10.9 ± 0.6	11.7 ± 1.3	14.2 ± 1.2	11.2 ± 0.7	12.7 ± 1.8
PCSA (cm2)	55.4 ± 8.6	40.1 ± 4.6	47.4 ± 10.2	37.8 ± 4.2	28.4 ± 5.1	32.9 ± 6.6

(c). Relationship between neural drive and muscle physiological cross-sectional area

There was a large significant correlation (r = 0.68, p = 0.0008; figure 3) between the ratio of VL/VM EMG amplitude and the ratio of VL/VM PCSA during the isometric knee extensions performed at 20% of MVC. There was a moderate correlation (r = 0.43, p = 0.051; figure 3) between these variables for the isometric knee extensions performed at 50% of MVC, although this relationship narrowly missed the threshold for statistical significance.

4. Discussion

We aimed to determine the relationship between the neural drive measured during isometric tasks and the force-generating capacity of VL and VM. This study has two major novel findings. First, participants used individualized strategies of coordination between VL and VM to perform the isometric knee extension tasks, and these strategies were robust between days. Second, there was a large (20% of MVC) or moderate (50% of MVC) correlation between the ratio of VL/(VL + VM) EMG amplitude and the ratio of VL/(VL + VM) PCSA, indicating that drive was biased by force-generating capacity; the greater the force-generating capacity of VL compared with VM, the stronger bias of drive to the VL. This leads to an imbalance of individual muscle force between these synergist muscles, the magnitude of which varies greatly between participants. These results are crucial to improve our current understanding of the complex interplay of individual muscle forces in the development of successful coordination strategies.

(a). Between-day reliability of neural drive

The inter-day reliability of the normalized VL and VM RMS EMG amplitude was good to excellent (table 1) indicating that the participants adopted a robust coordination strategy between these synergist muscles. Notably, reliability was less for VM (albeit remaining good) than for VL. Although we cannot rule out the possible influence of methodological issues such as crosstalk and amplitude cancellation, we do not believe that it would be significantly different between VM and VL. To support our assertion, a study that used elastography, which is insensitive to both crosstalk and amplitude cancellation, suggested a more variable force production in VM than VL across repeated isometric knee extensions [31]. It is, therefore, likely that this variability has a neurophysiological basis that might be related to the fact that in addition to knee extension, VM contributes to the control of the patellofemoral joint.

(b). Neural drive is not balanced between vastus lateralis and vastus medialis

Whatever muscles are under consideration, most studies that focus on coordination strategies during motor tasks report values averaged over a group of participants. However, this commonly used methodology conceals possible variability between participants [32,33]. In this way, the ratio between synergist muscles, such as the VL and VM, is often reported as close to 50%. This implies an equal contribution of synergist muscles to a given submaximal task and was the case for the current investigation; mean VL/(VL + VM) EMG ratio was 53.5 and 51.2% for knee extension performed at 20 and 50% of MVC, respectively. Although the neural drive was not significantly different between these muscles at the group level, individual data revealed a wide range of VL/(VL + VM) EMG ratios; an almost equal number of participants demonstrated greater VL EMG and greater VM EMG (e.g. range of VL/(VL + VM) EMG ratio: 33.6–74.7% and 42.2–64.7% at 20 and 50% of MVC, respectively). Pal et al. [10] reported similar between-subject variability during walking (ratio calculated as VL/VM ranged between 0.8 and 3). Although the VL and VM muscles might share most of their drive, they also receive muscle-specific neural drive [34]. Interestingly, the amount of muscle-specific neural drive varies between participants [34], which might explain the between-subject variability in VL/(VL + VM) EMG ratios.

Of note, the range of ratios observed in this study was smaller, and the group mean was closer to 50%, when the knee extensions were performed at 50% of MVC compared with those performed at 20% of MVC. This is consistent with previous observations made by Pincivero and Cohen [35], where VL and VM activation converged at near-maximal force levels, despite differences at submaximal levels. A convergence of the activation levels of VL and VM at higher levels of contraction is logical, as intensities close to MVC require the complete activation of all synergist muscles. Within the current experiment, the convergence of activation levels is highlighted by the near full voluntary activation of the quadriceps (and thus of both VL and VM) during MVCs as determined by the twitch interpolation technique. Overall, our results demonstrate an imbalance of neural drive between synergist muscles that differs between participants, and that this imbalance is more likely to occur at lower contraction intensities. This is important as lower contraction intensities are common during daily life activities.

(c). Muscle force-generating capacity is not balanced between vastus lateralis and vastus medialis

We considered the ratio of PCSA to represent the balance of muscle force-generating capacity between VL and VM. PSCA considers both fascicle length and muscle volume. Previous data on the PSCA of VM is lacking because of technical limitations (reviewed in [25]). In particular, muscle fascicles within the distal part of VM are arranged relatively parallel to the skin surface, and are longer than the width of conventional 2D ultrasound transducers. Therefore, the whole VM muscle fascicle cannot be visualized or estimated as is possible for muscles with a pennate arrangement, such as VL. Taking advantage of freehand 3D ultrasound, our study is one of the first to estimate the PCSA of VM in vivo. We have shown that VM PCSA is systematically smaller than VL PCSA. Similar to the VL/(VL + VM) neural drive ratios, a wide range of VL/(VL + VM) PCSA ratios were observed between participants (range 52.9–64.3%). This inter-subject variability in PCSA concurs with data reported for four adolescent girls using diffusor tension MRI (ratio calculated as VL/VM ranged between 1.1 and 3.1; average = 2.1) [36]. Similar variability was also reported in a study that measured cadaveric muscle volumes from 12 individuals (range: 47.3–68.3%, average: 61.5%) [15]. Until now it was unknown if or how this imbalance of force-generating capacity between synergist muscles relates to the distribution of neural drive.

(d). The relationship between neural drive and muscle force-generating capacity

Many possible combinations of VL and VM activation may produce a given submaximal knee extension torque. For example, it is possible that the nervous system might balance the force between synergist muscles of differing force-generating capacity by driving the muscle with the lower force-generating capacity at a higher level. However, this was not observed in the present study. Rather, we found a large (20% of MVC) or moderate (50% of MVC) positive linear correlation between the ratio of VL/(VL + VM) EMG amplitude and the ratio of VL/(VL + VM) PCSA, which indicates that neural drive is biased by force-generating capacity; the greater the force-generating capacity of VL compared with VM, the stronger bias of drive to the VL. This strategy seems logical considering the optimal control theory [37], which proposes that motor patterns are selected from many possibilities ensuring that movement costs (e.g. control, energetic, mechanical) are constantly minimized. However, our results may also support an alternative theory, the good enough theory, which proposes that a hierarchy of sensorimotor networks gradually adapt through trial-and-error learning to produce effective movements which are good enough to achieve the task goal [38]. Despite providing strong evidence of a positive relationship between neural drive and muscle PCSA, our results cannot explain the origin of this relationship. For example, it is possible that an individual's muscle morphology and architecture underlies the relationship, such that the nervous system develops/adapts optimally to bias drive to the muscle with larger PCSA. Alternatively, good enough neural coordination strategies may underpin muscle morphology and architecture, such that the muscle with greater drive leads to biased development of PCSA. However, as some participants distributed less drive to VL than VM (especially at 20% of MVC) and all participants had larger VL PSCA, muscle morphology/architecture is unlikely to depend on neural drive alone, unless there is a ‘ceiling’ to the development of VM. Finally, it is important to consider that the positive correlation we report might not indicate a causal relationship. Regardless of the basis of the correlation between neural drive and muscle force-generating capacity, our results provide strong evidence of an imbalance of force produced by VL and VM that varies considerably between participants.

(e). Balance of force between vastus lateralis and vastus medialis

If the product of normalized EMG amplitude with PCSA is considered as an index of force (arbitrary units), the ratio of this index between VL and VM ranges from 38.4 to 84.0% (mean: 61.4 ± 12.3%) and 48.8 to 71.0% (mean: 59.9 ± 7.5%) for knee extensions performed at 20 and 50% of MVC, respectively (figure 4). This highlights a very large variability between participants. However, beyond the intensity of the neural drive and muscle PCSA, individual muscle force depends on a combination of other biomechanical factors. Among them, the most important to consider for the comparison of VL and VM force is the specific tension (defined as maximal force per unit area) [39]. In vivo estimation of specific tension is challenging because there is no experimental technique to measure the force produced by an individual muscle [40]. Although specific tension is not expected to vary greatly between muscles with similar typology [41], VL exhibits a slightly higher percentage of type II fibres (approx. 60%) than VM (approx. 46%) [42]. As a higher percentage of type II fibres is thought to be associated with higher specific tension [43], the difference in muscle typology (albeit small) between VL and VM should accentuate the already large imbalance of force produced by these muscles. This has potential importance for musculoskeletal pain conditions that involve the patella, such as patellofemoral pain, which has been argued to relate to the imbalance of force between VL and VM [16].

Figure 4.

(a,b) Balance of force between the vastus lateralis (VL) and vastus medialis (VM) muscles. The index of force is calculated as the product of normalized electromyography (EMG) amplitude with physiological cross-sectional area (PCSA). These data highlight a high inter-individual variability with the majority of participants exhibiting more force produced by VL than VM. (Online version in colour.)

(f). Limitations

This experiment requires consideration of several methodological aspects. First, consistent with most previous studies, muscle fascicle length was measured at rest (for review, see [25]). This is because we used 3D ultrasound that requires a scan time of approximately 15 s during which motion of the muscle/limb must be minimized. It is important to consider that because muscle fascicle length decreases during muscle contraction, the muscle's PCSA increases [44]. We do not believe that the increase in PCSA during contraction will have influenced the conclusion of our study. This is because during the submaximal contractions the relatively higher activation of one muscle (VL or VM, depending on the participants) would be associated with relatively greater shortening of the fascicle length and consequently with relatively greater increases in PCSA of that muscle. Consequently, the correlation between the ratio of EMG and the ratio of PCSA would be strengthened further.

Second, the estimation of neural drive provided by surface EMG may be affected by physiological and non-physiological factors (reviewed in [45]). The most important to consider in our experiment are crosstalk, amplitude cancellation, and spatial variability of motor unit recruitment. First, crosstalk was limited by following the SENIAM recommendations and by checking the appropriate electrode location using ultrasound (i.e. fascicle direction and muscle borders). Second, EMG amplitude was normalized to that recorded during MVC. This normalization procedure has been shown to reduce the effect of amplitude cancellation on the EMG signals [46]. Finally, as surface EMG represents the algebraic summation of the motor unit action potentials under the recording electrodes, it is possible that preferential recruitment of superficial motor units for VL at low contraction intensities may explain, at least in part, the imbalance of EMG amplitude between VL and VM. To circumvent this limitation inherent to the EMG technique, future studies should consider the use of elastography because it may provide a better estimate of the normalized force [40].

Third, we did not measure the moment arm of VL and VM for either knee extension or patella motion (tilt and rotation), which makes it difficult to further interpret the biomechanical effects of the imbalance of force between these muscles. However, the moment arm relative to the midpoint of the tibiofemoral flexion/extension axis is not different between VL and VM [17,18], particularly when the knee is flexed [47], as is the case in this study. Consequently, we argue that it is unlikely that the imbalance of force is counterbalanced by different moment arms.

5. Conclusion

This study provides new insight into the strategy for force sharing between synergist muscles and provides evidence that drive is biased by force-generating capacity; the greater the force-generating capacity of VL compared with VM, the stronger the bias of drive to the VL. Although this is efficient in some respects (e.g. minimizes the neural drive), this strategy is very likely to result in force (and also torque) imbalance between the synergist muscles. Further research is needed to determine the potential role of this imbalance in the completion of successful coordination strategies/movements and its long-term effects, particularly in the occurrence/persistence of musculoskeletal conditions.

Ethics

Participants provided informed written consent. The ethics committee of The University of Queensland approved the study and all procedures adhered to the Declaration of Helsinki.

Data accessibility

Data deposited in Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.h3b27.

Authors' contributions

Conception and design of the experiments: F.H., P.W.H., and K.T.; collection, analysis, and interpretation of data: F.H., C.G., D.B., B.R.; drafting the article or revising it for important intellectual content: F.H., C.G., B.R., P.W.H., and K.T. All authors approved the final version of the manuscript.

Competing interests

We declare we have no conflict of interest.

Funding

NHMRC provide research fellowships for P.H.: ID401599 and K.T.: ID1009410. Project support was provided by an NHMRC program grant (P.H.: ID631717), Center of Advanced Imaging, University of Queensland (project support ID15003) and Région Pays de la Loire (QUETE).