Mapping of spatial and temporal heterogeneity of plantar flexor muscle activity during isometric contraction: correlation of velocity-encoded MRI with EMG

Abstract

The aim of this study was to assess the correlation between contraction-associated muscle kinematics as measured by velocity-encoded phase-contrast (VE-PC) magnetic resonance imaging (MRI) and activity recorded via electromyography (EMG), and to construct a detailed three-dimensional (3-D) map of the contractile behavior of the triceps surae complex from the MRI data. Ten axial-plane VE-PC MRI slices of the triceps surae and EMG data were acquired during submaximal isometric contractions in 10 subjects. MRI images were analyzed to yield the degree of contraction-associated muscle displacement on a voxel-by-voxel basis and determine the heterogeneity of muscle movement within and between slices. Correlational analyses were performed to determine the agreement between EMG data and displacements. Pearson's coefficients demonstrated good agreement (0.84 < r < 0.88) between EMG data and displacements. Comparison between different slices in the gastrocnemius muscle revealed significant heterogeneity in displacement values both in-plane and along the cranio-caudal axis, with highest values in the mid-muscle regions. By contrast, no significant differences between muscle regions were found in the soleus muscle. Substantial differences among displacements were also observed within slices, with those in static areas being only 17–39% (maximum) of those in the most mobile muscle regions. The good agreement between EMG data and displacements suggests that VE-PC MRI may be used as a noninvasive, high-resolution technique for quantifying and modeling muscle activity over the entire 3-D volume of muscle groups. Application to the triceps surae complex revealed substantial heterogeneity of contraction-associated muscle motion both within slices and between different cranio-caudal positions.

skeletal muscles are highly heterogeneous tissues that are characterized by regional variations in muscle architecture (5, 25, 28, 57) and histochemical fiber type composition (1, 33, 51). This marked nonuniformity allows for the organization of muscles in subunits that are sometimes referred to as neuromuscular compartments (59) or task groups of motor units (23). Principles of movement economy, such as the Henneman size principle (19), dictate that the recruitment of such task groups would depend on the force demands as well as the velocity and excursion of certain movements and suggest that, in submaximal contractions, muscles would not be uniformly activated across their entire bellies.

The concept of muscles within muscles that may be independently controlled by the central nervous system (68) has received empirical support by a number of studies that have investigated intramuscular differences in task-specific activation patterns by electromyography (EMG) with either multichannel surface (24, 27, 41, 65) or indwelling electrodes (7, 45, 69). However, when attempting to assess muscle activation patterns with high spatial resolution in an entire muscle, the applicability of both of these techniques is limited. This is because surface EMG (sEMG) allows only measurement of the superimposed EMG activity of larger, superficial muscle areas, which complicates the decomposition of the recorded signals into the contributions of single-task groups of motor units (14), whereas intramuscular EMG (iEMG) suffers from its very small area of detection. Also, the placement of indwelling electrodes is invasive, requiring appropriate sterile conditions and personnel expertise, and may cause discomfort, particularly during muscular contraction.

An alternative approach to measuring muscle activity is to use noninvasive medical imaging techniques that allow dynamic studies of muscle behavior during contraction. The measurement principles of EMG and medical imaging techniques are fundamentally different. EMG assesses a muscle's response (in the form of electrical activity) to incoming stimuli (action potentials) from motor neurons, and either represents an average of neural drives to multiple (sEMG), or fewer, or even single motor units (iEMG). By contrast, sonography (61) and specific magnetic resonance imaging (MRI) sequences enable the visualization, and thus quantitation, of contraction-associated muscular strains. One such MRI technique is represented by velocity-encoded phase-contrast (VE-PC) MRI, which provides information about the three-dimensional (3-D) displacement of tissues on a voxel-wise basis and, therefore, facilitates measurements with the greatest in-plane resolution while covering large areas of interest. Our laboratory has specialized in the development and refinement of modalities to apply VE-PC MRI to the study of skeletal muscles and tendons (55, 56). Using this technique, we have made intriguing observations of contraction-related muscle behavior and were among the first to report nonuniformity of muscle displacement as measured at the muscle group (16), single muscle (10, 15, 29), and even the single fascicle level (54). Although it is intuitive to assume that the heterogeneity in muscle tissue movement would be reflected in neuromuscular recruitment patterns, to date, no data directly establishing a relationship between contraction-associated muscular electrical activity and tissue strains have been published.

On the basis of these considerations, the goals of the present study were twofold: 1) to assess the correlation between EMG-based and VE-PC MRI-based data obtained during isometric muscular action; and 2) to apply VE-PC MRI to an in-depth analysis of the contractile behavior of the human plantarflexor muscles, toward mapping, on a voxel-by-voxel basis over the entire volume of the triceps surae muscles, the spatial and temporal distribution of areas of muscle activation during submaximal isometric contraction. Previous studies investigating region-specific muscle architecture have demonstrated that a marked nonuniformity exists within the regions of the gastrocnemii muscles, with fascicle length being shortest in the most proximal regions and longest in the most distal regions (9, 57). In the soleus muscle, by contrast, regional differences in muscle architecture are less pronounced (9). Because muscle architecture is known to be one of the main determinants of muscle function (38, 47), we hypothesized that muscle tissue displacements would show a proximal-to-distal gradient in the gastrocenmius muscle, but be distributed more homogeneously in the soleus. To test this hypothesis, we studied a cohort of 10 young and healthy volunteers to obtain triceps surae EMG and axial-plane VE-PC cines at multiple cranio-caudal positions during submaximal isometric contraction.

MATERIALS AND METHODS

Subjects.

Ten young and healthy subjects (7 men, 3 women, age 27.0 ± 6.2 yr, height 169.6 ± 11.8 cm, weight 69.3 ± 13.7 kg) volunteered to participate in this study. To be included, participants had to be free of previous injuries to the lower legs and sign informed consent forms. Only the dominant leg, defined as the leg preferentially used to kick a ball (right leg for all subjects), was examined. The study was approved by the Institutional Review Board of the University of California–San Diego (project no. 140487) and conducted in agreement with the ethical principles for medical research outlined in the Declaration of Helsinki (70).

Experimental setup.

Before the MRI scans, maximum voluntary contraction (MVC) strength was determined during isometric plantarflexion contractions in all subjects. For this purpose, participants were instructed to lie down on the MRI examination bed in the supine position, with their hips and knees in a fully extended position. The tested leg was placed in the posterior half of a custom-made fiberglass cast (Techform Casting Tape; Össur, Foothill Ranch, CA), which was tightly fixed to the lower leg with elastic bandages to fix the ankle at a 90° angle. An additional Velcro strap was wrapped around the ankle to minimize inadvertent joint rotation during contraction. Leg and cast were then positioned and fixed inside a standard MRI head coil. To measure plantarflexion strength, an MRI-compatible optical pressure transducer (Luna Innovations, Roanoke, VA) was glued onto the sole of the cast. The pressure exerted during plantarflexion contractions was detected by this transducer, transmitted via a fiber-optical cable, voltage-converted by a spectrometer (Fiberscan; Luna Innovations) stationed in the console room, and digitized at a sampling frequency of 200 Hz using a standard A/D converter (NI USB-6008; National Instruments, Austin, TX). An in-house LabVIEW program (NI LabVIEW 2011; National Instruments, Austin, TX) was used to record and store the signal onto a laptop computer. Plantarflexion MVC strength was determined during three consecutive isometric contractions, each ∼5 s long and interspersed by ∼1 min of passive recovery. The voltage recorded during the best of these trials was used to set the target intensity for the subsequent submaximal contractions to be performed during VE-PC MRI scans. This value was equivalent to 40% of the individual MVC strength, which we found to represent the maximum intensity that subjects could reproduce consistently over the entire duration of the scanning protocol while avoiding undue motion artifacts or incurring fatigue. The experimental setup for the acquisition of images is depicted schematically in Figure 1 and described in detail by Sinha et al. (55). In brief, VE-PC scans (see MRI protocol) require the execution of ∼70 contraction-relaxation cycles (phase-encoding repetitions). Because consistency of movement over these cycles is paramount for image quality, a second LabVIEW program was also used to project, onto the magnet face, a rectified sine wave, which served as visual guide to participants in terms of both timing and contraction strength. The actual force exerted by the subject was also projected synchronously so that the subject was able to gauge the degree of fidelity with which he or she was able to follow the visual cue, and incorporate any necessary corrections to his contractions. To synchronize the acquisition of images with the contraction-relaxation cycles, each phase-encoding repetition was gated by a trigger obtained by simultaneously differentiating the voltage curve output (by a third LabView program) by the spectrometer, to generate an electrocardiogram-like trigger signal, which was then entered into the scanner electrocardiogram input. After completion of the MRI protocol, the examination bed was undocked from the scanner and transferred into the console room to obtain EMG recordings (see EMG analyses). Participants then performed yet another series of MVC tests and subsequent 70 submaximal contractions under the same experimental conditions as within the MRI scanner. Care was always taken to avoid fatigue on part of the subjects.

Fig. 1.

Illustration of the experimental setup for velocity-encoded phase-contrast (VE-PC) image acquisition. Subjects were instructed to execute plantarflexion contraction-relaxation cycles (∼70) by rhythmically pushing against the sole of a cast fixed to their lower legs. The pressure exerted was detected by an optical Fabry-Perot interferometer-based pressure transducer and transmitted via a fiber optical cable to a spectrometer stationed outside the magnet room. The voltage changes output by this device were 1) differentiated to create an electrocardiogram-like trigger signal that was fed into the scanner to gate the acquisition of images with respect to the contraction-relaxation cycles; and 2) recorded by a custom-made LabVIEW program. The recorded signal, together with a half-wave rectified sine wave serving as a visual cue, was projected onto the magnet face to facilitate the subject executing more consistent force patterns by gauging each exerted contraction in terms of timing and strength.

MRI protocol.

MRI scans were performed on a 1.5 T magnetic resonance scanner (Signa version 12; General Electric Medical Systems, Milwaukee, WI) using a standard transmit/receive birdcage head coil, which allowed for the entire triceps surae complex to be covered without the need to reposition subject or coil. The imaging protocol consisted of a three-plane localizer, which was used to identify the most proximal position of the gastrocnemii muscles and the most distal position at which the soleus muscle inserts into the free tendon. Subsequently, 10 equally spaced axial-plane VE-PC scans were obtained at positions coinciding with 10–90% (steps of 9%) of triceps surae length (Fig. 2). Interslice gaps were therefore dependent on the individual muscle length and ranged between 25 and 36 mm (mean 31 ± 3 mm). The imaging parameters for VE-PC scans were as follows: echo time, 6.3 ms; repetition time, 24.2 ms; no. of excitations, 2; flip angle, 20°; slice thickness, 5 mm; field of view, 20 × 14 cm; 256 × 256 matrix; and 10 cm/s 3-D velocity encoding. With four views per segment, this resulted in a typical total of 70 contractions to be exerted by the subject for VE-PC image acquisition. The sequence produces one set of morphological (magnitude) images and three sets of VE-PC images, each consisting of 60 (view-shared) phases (cine) acquired during the approximately 3-s contraction-relaxation cycle. The three VE-PC cines result from the velocity encoding in the x-, y-, and z-directions and use a gray scale to reflect the directional velocity of tissue movement on a voxel-wise basis.

Fig. 2.

Schematic showing magnetic resonance imaging (MRI) slice locations and demonstrative VE-PC images as obtained at rest and during muscular contraction.

MRI postprocessing and image analyses.

The set of magnitude images from the sequence described above was used to outline the contours of the medial and lateral gastrocnemius (GM and GL, respectively) and soleus (SOL) muscles. This served to generate masks and report muscle-specific displacement data. The velocity-encoded images were first corrected for shading artifacts arising from magnetic field inhomogeneities and chemical shifts (55), and then used to track the displacement of each voxel of muscle tissue on a frame-by-frame basis throughout the contraction-relaxation cycle. For this purpose, the images reflecting the velocity in the x- and y-directions were evaluated to track the in-plane displacement of each volume element. Accounting for the third velocity component, the magnitude of total displacement with respect to the original position in the first frame was then calculated as the square root of the sum of squared displacements in each direction [DISP = √(x² + y² + z²)]. Note that through-plane tissue movements (i.e., in the cranio-caudal direction) could not be tracked due to the 2-D nature of slice acquisition. Therefore, uniformity of velocities in this axis had to be assumed. Blood vessels were identified algorithmically rather than by a labor-intensive manual process, and were excluded from further analysis if intramuscular regions were characterized by a consistent increase in tissue displacement (i.e., volume elements moving steadily farther away from their original position, as opposed to muscle tissue points that were found to return to their original locations on the completion of a cycle). Prominent blood vessels identified by the algorithm were validated through visual comparison by an expert radiologist. The displacements of the remaining voxels were averaged over the cross-sectional areas of the GM, GL, and SOL muscles to obtain mean curves of muscle movement for comparison with EMG data. To assess the regional heterogeneity of muscle activity, the muscle-specific maximum displacements occurring during the contraction-relaxation cycle were separately extracted for each slice, expressed as a percentage of the maximum displacement measured in the whole respective muscle, and grouped for statistical comparison of proximal (GM and GL, 10–19% of triceps surae length; SOL, 19–37%), middle (GM and GL, 28–37%; SOL, 46–63%), and distal (GM and GL, 46–55%; SOL, 72–90%) muscle regions. The same procedure was applied for comparison of the times at which the maximum displacements were observed. For in-plane heterogeneity (within slices), the voxel-specific maximum total displacements and the respective times were determined, filtered using a 2-D anisotropic diffusion filter (52), and visualized on color maps of muscle movement.

EMG measurements and further data analyses.

During isometric contractions, the sEMG activity of the GM, GL, and SOL muscles was recorded at a sampling frequency of 1 kHz using an 8-channel EMG system (Myosystem 1200; Noraxon, Scottsdale, AZ). Skin preparation and electrode placement followed the general recommendations for sEMG outlined in the SENIAM guidelines (20). Specifically, two bipolar Ag-AgCl surface electrodes (Blue Sensor N; Ambu, Ballerup, Denmark) were placed over the most prominent bulges of the GM and GL (typically at ∼35% of the length of an imaginary line from the popliteal fossa to the calcaneus at the point of insertion of the Achilles tendon) along the central line of the muscles. For the SOL, electrodes were placed in the distal half of the muscle (below the distal end of the GM) and medially to the muscle's central line. Care was taken to align the electrodes perpendicularly to the expected orientation of muscle fibers. The raw EMG data were preamplified (gain ×1,000, CMRR >100 dB), digitized, and band-pass filtered between 10 and 500 Hz. The signal was further rectified and smoothed by taking the root mean square of the signal (RMS-EMG) using a time window of 300 ms. The nature of our movement paradigm consisting of slow (3-s cycle duration), isometric contractions allowed for this rather long time window to be chosen; assuming a relatively constant EMG activity, longer smoothing window lengths significantly increase the signal-to-noise ratio of EMG recordings (58). The measurements obtained during MVC trials were integrated over 500 ms around the peak torque and used for later MVC normalization of the recordings from submaximal contractions. Analyses of EMG data were performed using commercially available software (Myoresearch XP; Noraxon, Scottsdale, AZ). To synchronize force and EMG measurements, the trigger signal first used to gate the image acquisition was then entered into and recorded by both the EMG system and the LabView-based program for force measurements. Using custom-made MATLAB routines (MATLAB R2012b; Mathworks, Natick, MA), the recorded signals were phase-aligned and averaged over the 70 contraction-relaxation cycles to obtain curves of mean force and EMG activity. Threshold-based MATLAB algorithms were applied to automatically detect the onset and offset, and the maxima of both EMG and VE-PC recordings, and report the corresponding times with respect to the onset and maximum of force production. Further parameters extracted to test the agreement of EMG and VE-PC data included the total activity duration as well as the slope, defined as the rate of relative rise of these signals (normalized to the respective maxima) within 0.5 s after onset. To better illustrate the data extraction procedures, demonstrative data curves are shown in Figure 3.

Fig. 3.

Demonstrative force-electromyography (EMG) and force-displacement (VE-PC MRI) curves illustrating the data extraction procedures. Disp indicates the sum of muscular displacements as measured by velocity-encoded phase contrast MRI in the center of the muscle bellies.

Statistical analyses.

After testing data for distributional normality using Shapiro-Wilks tests, Pearson's coefficients were calculated to assess the correlation of EMG and VE-PC recordings. Independent sample t-tests were used to test for differences between results obtained with the two different techniques. Because significant Shapiro-Wilk tests indicated a violation of the assumption of normality (P < 0.05), Kruskal-Wallis tests were applied to assess differences between results from multiple muscle regions. Post hoc Bonferroni-adjusted Mann-Whitney U-tests were used where appropriate. Results are reported as means ± SD, and the level of statistical significance was set at α = 0.05. All tests were performed using SPSS for Mac OSX (SPSS 22.0; SPSS, Chicago, IL).

RESULTS

Agreement between sEMG and VE-PC recordings.

Pearson's coefficients were calculated to test the correlation of sEMG recordings and VE-PC-based sums of muscular displacements (DISP) as measured in the center of the muscle bellies (i.e., at the single cranio-caudal slice position in closest proximity to the respective sEMG electrode). These tests demonstrated good agreement between sEMG and DISP in all heads of the triceps surae complex (GM r = 0.87 ± 0.10, GL r = 0.88 ± 0.08, SOL r = 0.84 ± 0.09, P < 0.001). To further examine the congruence of sEMG and DISP, onset times at which sEMG and DISP signals indicated a rise in muscular activity and total activity durations were compared. With respect to the first, DISP was found to consistently trail sEMG, although by a small amount (0.06 to 0.09 s). These differences in onset times were significant in all heads of the triceps surae complex. However, the rise in DISP still preceded the rise in force by 0.14 to 0.15 s. Total activity durations as measured by sEMG were shorter by 0.02 (SOL) to 0.07 s (GM and GL), although these differences failed to reach significance. Compared with DISP, sEMG signals demonstrated a significantly steeper relative incline in the first 0.5 s after onset with the differences in slopes ranging between 0.31 and 0.38 (P < 0.05). Accordingly, DISP curves reached their maxima significantly later than sEMG signals by 0.04 to 0.12 s, however, still preceding the time of maximum force generation by 0.14 to 0.20 s. All data and statistical tests reflecting the agreement of EMG and VE-PC-based recordings of muscle activity are summarized in Table 1.

Table 1.

Congruence of EMG and VE-PC-based recordings of muscle activity

	EMG		VE-PC
	Mean	SD	Mean	SD	P
GM onset tme, s	−0.23	0.06	−0.15	0.06	0.008*
GM activity duration, s	1.92	0.21	1.99	0.19	0.476
GM slope	1.55	0.23	1.17	0.22	0.002*
GM maximum time, s	−0.28	0.11	−0.19	0.12	0.127
GL onset time, s	−0.22	0.10	−0.15	0.07	0.143
GL activity duration, s	1.85	0.27	1.92	0.15	0.473
GL slope	1.54	0.20	1.21	0.27	0.006*
GL maximum time, s	−0.24	0.09	−0.20	0.13	0.436
SOL onset time, s	−0.23	0.11	−0.14	0.11	0.080
SOL activity duration, s	1.96	0.27	1.98	0.17	0.822
SOL slope	1.53	0.23	1.22	0.39	0.043*
SOL maximum time, s	−0.26	0.12	−0.14	0.07	0.018*

EMG, electromyography; GL, lateral gastrocnemius muscle; GM, medial gastrocnemius muscle; SOL, soleus muscle; VE-PC, velocity encoded-phase contrast (MRI). Onset and maximum times are reported relative to the onset and maximum of force, respectively.

^*Significant difference.

Regional and temporal heterogeneity in muscle activity.

Comparison of the maximum muscle-specific displacements between proximal, middle, and distal muscle regions indicated that, compared with distal and proximal regions, displacements were larger by 10–25% in the middle of both GM and GL muscles. Kruskal-Wallis tests revealed that these differences were significant in both GM (χ² = 9.052, df = 2, P = 0.011) and GL (χ² = 7.885, df = 2, P = 0.019). Bonferroni-adjusted Mann-Whitney U-tests performed to follow up these results showed statistical differences between the middle and proximal regions in GM (U = 7.500, P = 0.006), and between the middle and distal regions in GL (U = 8.000, P = 0.030). In SOL, by contrast, differences in displacements between muscle regions failed to reach significance (χ² = 2.327, df = 2, n.s.). Descriptive statistics of the muscle region-specific maximum displacements are shown in the boxplots in Figure 4. For times at which displacements reached their maxima, no significant differences were found between regions in GM (χ² = 0.067, df = 2, n.s.), GL (χ² = 3.680, df = 2, n.s.), and SOL (χ² = 4.452, df = 2, n.s.).

Fig. 4.

Box plots showing the contraction-associated displacements (VE-PC MRI) of muscle tissue in proximal (prox), middle (mid), and distal (dist) regions. GM, medial gastrocnemius muscle; GL, lateral gastrocnemius muscle; SOL, soleus muscle.

Differences in maximum tissue displacements were observed not only between different cranio-caudal regions, but also in-plane (i.e., within the same slice, as is evident from the demonstrative color maps in Figure 5A). The heterogeneity of displacements is reflected by statistical measures of dispersion: average muscle movements amounted to 0.39–0.51 cm with standard deviations ranging between 0.10 and 0.17 cm. To further quantify the in-plane heterogeneity, we determined the average displacements in areas of 5×5 pixels around the single pixel with the greatest (so-called hot) and smallest (so-called cold) regions of tissue displacement. The comparison of results revealed significant differences in all slices (P < 0.01). More specifically, the average displacements in cold regions amounted to only 16.6% to 38.8% of those measured in hot regions.

Fig. 5.

Demonstrative VE-PC magnetic resonance images showing the heterogeneity of maximum muscular displacements observed during submaximal isometric contraction. A: cross-sectional images obtained at different cranio-caudal positions reflecting the in-plane heterogeneity. Red (i.e., hot) indicates areas with large muscle displacement; blue (i.e., cold) areas are relatively static. B: three-dimensional reconstruction of mostly active (maximum displacements ≥0.5 cm) muscle regions. Areas exceeding a total displacement of 0.5 cm were identified in each individual slice and used to create volumetric information through forward projection.

DISCUSSION

The goals of the present study were to obtain EMG and VE-PC MRI data reflecting muscular activity, to test the agreement of the results from two disparate modalities, and to further use VE-PC MRI to assess, over the entire volume of all heads of the triceps surae on a voxel-by-voxel basis, the intramuscular heterogeneity of muscle activity associated with submaximal, isometric contraction. Comparison of results demonstrated good agreement (0.84 < r < 0.88) between EMG activity and the magnitude of muscle tissue displacement, suggesting that VE-PC MRI may provide a surrogate technique for studying muscle activity in large areas of interest and modeling it in 3-D. Applying this technique to the study of the plantarflexor muscles, we further found significant differences in muscular activity both between different cranio-caudal positions in the GM and GL, and in-plane, in all heads of the triceps surae complex.

Measurements of skeletal muscle activity are routinely performed in both biomechanical research and clinical settings to investigate the forces generated by muscles, study the motor control of movements, perform motion analyses, and detect neuromuscular abnormalities. To assess exercise-induced muscle activation patterns, sEMG techniques have been most commonly used. The electrical activity detected by sEMG represents the muscle's response to stimulation by supporting motor neurons. Because the electrodes detect the superimposed electromyographic activity of a large number of motor units, sEMG allows only for statements about relatively large muscular areas and is therefore not considered to be specific enough to target small regions of interest. Other shortcomings include the possibility of crosstalk and signal decay at greater distances from the recording electrode (14). Intramuscular EMG, by contrast, allows for recordings with a higher spatial resolution, is less susceptible to artifacts associated with movement and crosstalk from nearby muscles, and may be used to study muscles located deep within the body (44). However, the applicability of this technique is limited by its invasiveness and the small area of detection of indwelling electrodes.

In view of the shortcomings outlined above, several alternative techniques have been developed and proposed as noninvasive surrogate measures of muscle activity. These include 1) MRI-based measurements of T2 values reflecting the exercise-induced and osmotically driven water shifts from extracellular to intracellular spaces (11, 30); 2) magnetic resonance spectroscopy (4, 6, 17, 34, 67) or chemical shift imaging (21) using either ¹H or ³¹P nuclei to track changes in muscle metabolism; 3) measurements of muscle regional perfusion and oxygenation, which are closely coupled to muscle activity (12, 60, 66); 4) more recent approaches such as magnetic resonance elastography, which infers activity on the basis of tissue stiffness measurements (18); and 5) tracer-based techniques to measure muscle glucose uptake (50). All of these techniques are limited by their relatively low spatial resolution and the indirectness of the approach. VE-PC MRI, in contrast, yields instantaneous direct measures of contraction-associated muscle movement, and therefore activity, in large regions of interest and with the greatest spatial resolution of existing methodologies, and may therefore circumvent many of the limitations of the aforementioned techniques.

In this study, we found that RMS-rectified sEMG curves correlated highly with the magnitude of muscular displacements as measured near the recording EMG electrodes (0.84 < r < 0.88). The good agreement of measurements is further evidenced by the small differences in total activity durations (0.02 to 0.07 s) and the acceptable congruence of onset times. With regard to the latter, the rise in the signal reflecting muscular displacements was found to consistently trail the sEMG recordings by 0.06 to 0.09 s while still preceding the onset of force generation by 0.14 to 0.15 s. In part, this time lag between the increase in electrical activity and the mechanical response of the muscle may be ascribed to the time duration of several physiological processes involved in the generation of force in the skeletal muscle. These include the conduction of the action potential along the T-tubule system, the release of calcium from the sarcoplasmic reticulum, the cross-bridge formation between myofilaments, and the tautening of in-series elastic components (8). However, the degree of electromechanical delay measured in this study was substantially larger than that typically observed in the human plantarflexor muscles (46, 53). A possible explanation for the lack of synchrony between the onset of EMG activity and the first indication of contractile activity lies in the nature of our movement paradigm. In cyclical contraction-relaxation cycles, the associated rhythmical stretch and recoil of tendons may strongly influence the neural control strategy of movement and result in muscle excitation substantially preceding force generation (39, 49). In addition, the temporal resolution of VE-PC scans, determined by the number of phases acquired during the contraction-relaxation cycle (here, 0.05 s⁻¹), is substantially lower than that of sEMG recordings (here, 0.001 s⁻¹). Therefore, VE-PC MRI is suboptimal for the accurate determination of the timing of muscular activity.

One of the strongest advantages of the VE-PC technique is its ability to scan over large areas of interest, thereby facilitating the simultaneous study of intramuscular and intermuscular heterogeneity of activity in entire muscles or even synergistic muscle groups. In our study, significant Kruskal-Wallis test results revealed statistical differences in the maximum displacements in different cranio-caudal regions of both GM and GL, with the largest tissue movements observed in the center of the respective muscle bellies. It is important to realize the differences between muscular activity as determined by sEMG and VE-PC. Whereas sEMG measures the electrical potential generated by neurologically activated muscle fibers, VE-PC measures the resulting velocity and magnitude of muscular shortening and associated overall displacement. This downstream parameter will be affected not only by neural activation but by a manifold of muscle parameters, including regional differences in fiber type composition (36); muscle architecture (2); the region-specific material properties of the muscle's extracellular matrix (22), which allows for the lateral transmission of force and displacement; and the history of previous muscle action (26). Visualization of the timing and degree of muscular shortening in large areas of interest, as facilitated by VE-PC, reflects the interdependent influence of all of these factors and may therefore substantially contribute to the understanding of muscular force generation and gradation. Of particular note, it has also been demonstrated that in response to long-term resistance training, many muscles hypertrophy in a nonuniform manner (43, 48), and that regions evidencing greater muscle growth may coincide with those more active during exercise (63). Hence, accurate models of muscle tissue displacement might even help predict the consequences of chronic loading. The main limitations of the VE-PC technique compared with EMG are the substantially higher cost of MRI and the limited availability of the scanner and the algorithms to quantitatively analyze the results. It is also important to note that the spatial resolution along the proximo-distal axis is determined by slice thickness and interslice gaps because the 2-D nature of image acquisition does not allow for information about tissue movements to be quantified between slices. To cover large areas of interest, it is therefore necessary to either increase the number of cross-sectional images acquired (which results in correspondingly longer scanning times), or choose larger interslice gaps, which reduces the accuracy of results when interpolating between slices. We note that spatial 3-D magnetic resonance acquisition protocols exist in which the velocity can also be encoded in all three directions, but given the large cranio-caudal extent of the leg, these 3-D scans would require a prohibitively long scan time.

In the gastrocnemii muscles, we found VE-PC displacements to be relatively largest in the center of the muscle belly, where muscle tissue movements were larger by approximately 28–29% than in more proximal (GM) or distal (GL) regions, respectively. In the SOL muscle, by contrast, displacements did not differ significantly between muscle regions. Considering results of previous studies (both cadaver- and imaging-based) investigating the region-specific muscle architecture within the gastrocnemii muscles, we had expected to find a proximal-to-distal gradient of tissue movement in these muscles. This is because pennation angles are greater and, consequently, fiber lengths are shorter near the origins of these muscles (9, 57). Shorter fibers consist of a smaller number of sarcomeres arranged in series so that the arrival of a single action potential would induce less total fiber shortening. Previous data obtained by us (54) and others (37) indeed confirm that GM fascicle length changes during submaximal contraction are smallest in the most proximal muscle region.

The present finding that muscle tissue displacements within the various regions of the gastrocnemii were actually largest in the middle (and not the most distal) regions may also be related to region-specific differences in neuromuscular activity. Multichannel sEMG studies have provided evidence that triceps surae recruitment patterns may vary both between subjects and in dependency of force requirements (59). Although the interpretation of EMG data obtained under different experimental conditions is complicated, Wakeling (64) reported that during slow-cadence, low-torque cycling, the EMG intensity of the GM showed a distal-to-proximal gradient. Similarly, Vieira et al. (62) used 2-D arrays of electrodes to localize areas of high activity within the gastrocnemii muscles. During isometric plantarflexion contraction, the highest activity as measured by EMG was observed in the proximal ends of the GL (note that although the 2-D grid offers the ability to localize activity on the surface, spatial information along the depth—i.e., the anterior-posterior axis of the muscle—is still not available). If similar intramuscular recruitment patterns could also be observed during low-force isometric contractions, the combined influence of region-specific fascicle lengths (favoring greater displacements in the distal muscle regions) and neuromuscular activity (potentially greater in proximal regions) might explain why, in our study, the overall muscle tissue movement was largest in the center of the muscle belly. Along the same line of argument, intramuscular differences in SOL architecture (9) and neuromuscular activity (64) have been reported to be less pronounced than in the gastrocnemius muscle, which might result in relatively homogeneous muscle tissue displacements. However, considering the manifold of factors potentially affecting VE-PC results, these explanations are likely to be overly simplistic.

In this study the directly measured muscle displacements represent the sum of displacements along all three spatial directions. Consequently, they will reflect the final displacement of tissue points that will be determined not only (and primarily) by muscle fiber shortening, but modifications thereof by various factors such as inert muscle or the extracellular matrix around it. Although the correlation of muscle displacements with EMG was good, such movements could also result from the deformation of a certain muscle due to co-contraction of its agonists or from principally inactive muscle regions being dragged along by contracting parts due to lateral transmission of force over the network represented by the muscle's extracellular matrix. Further complexity arises when considering potential regional specialization of muscles in terms of fiber type composition. Although we are not aware of comparable studies carried out with the human triceps surae complex, data from animal studies suggest that fibers of a given myosin heavy chain isoform may be clustered in certain muscle regions, and that these clusters vary along the muscle's proximo-distal length (32). It may be speculated that fatigue-resistant fibers, which are expected to be preferentially recruited for the execution of repeated low-force contractions, would lie predominantly in the middle of the gastrocnemii but be distributed more homogeneously in the (predominantly slow twitch) SOL muscle. Note that some measurement inaccuracy may stem from the cyclic nature of our movement paradigm. Repeated contractions have been shown to induce creep in tendinous tissues (35, 40), necessitating greater fiber shortening to generate the same amount of contractile force.

Our data further suggest that the heterogeneity of displacements was not restricted to only along the cranio-caudal direction, but it was also evident within the same axial-plane slices. In fact, in-plane comparison of muscular regions revealed that average displacements may differ by a factor of up to 6 between hot and cold zones. Jointly, this results in a complex 3-D pattern of muscle strain, as evidenced in the demonstrative plot in Figure 5B. In the future, sophisticated computational models developed by our group (3) will help integrate these 3-D architectural, compositional, motor activation, and other factors, to better predict the final tissue displacement and understand the complex nature of their interactions in generating muscle force.

In conclusion, we successfully applied VE-PC MRI to study contraction-related strain patterns in the human triceps surae complex. Correlation with sEMG data provided evidence that VE-PC may be used to reliably assess muscle activity in large areas of interest. Comparison between different muscle regions revealed that tissue displacements associated with contraction differ strongly both between different cranio-caudal positions and within single slices. This spatial distribution of strains is likely to be influenced by a large number of physiological parameters that may interact in a complex manner to generate and grade muscle forces.

GRANTS

This work was supported by National Institute of Arthritis and Musculoskeletal and Skin Diseases Grant 5RO1-AR-053343-08.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

R.C. and S.S. conception and design of research; R.C. and V.M. performed experiments; R.C. and V.M. analyzed data; R.C., U.S., and S.S. interpreted results of experiments; R.C. and V.M. prepared figures; R.C. drafted manuscript; R.C., V.M., U.S., and S.S. approved final version of manuscript; U.S. and S.S. edited and revised manuscript.