Vol. 87, Issue 6, 2107-2114, December 1999
Pixel T2 distribution in functional magnetic resonance images
of muscle
Barry M.
Prior1,
Jeanne M.
Foley1,2,
Roop C.
Jayaraman2, and
Ronald A.
Meyer1,3
Departments of 1 Physiology,
2 Kinesiology, and
3 Radiology, Michigan State
University, East Lansing, Michigan 48824
 |
ABSTRACT |
Increases
in skeletal muscle 1H-NMR
transverse relaxation time (T2) observed by magnetic resonance imaging
have been used to map whole muscle activity during exercise. Some
studies further suggest that intramuscular variations in T2 after
exercise can be used to map activity on a pixel-by-pixel basis by
defining an active T2 threshold and counting pixels that exceed the
threshold as "active muscle." This implies that motor units are
nonrandomly distributed across the muscle and, therefore, that the
distribution of pixel T2 values ought to be substantially broader after
moderate exercise than at rest or after more intense exercise, since
moderate-intensity exercise should recruit some motor units, and hence
some pixels, but not others. This study examined the distribution of
pixel T2 values in three muscles (quadriceps, anterior tibialis, and biceps/brachialis) of healthy subjects (5 men and 2 women, 18-46 yr old) at rest, after exercise to fatigue (50% 1 repetition maximum at 20/min to failure = Max), and at 1/2Max (25% 1 repetition
maximum, same number of repetitions as Max). Although for each muscle
there was a linear relationship between exercise intensity and mean pixel T2, there was no significant difference in the variance of pixel
T2 between 1/2Max and Max exercise. There was a modest (10-43%) increase in variance of pixel T2 after both exercises compared with rest, but this was consistent with a Monte Carlo simulation of muscle activity that assumed a random distribution of
motor unit territories across the muscle and a random distribution of
muscle cells within each motor unit's territory. In addition, 40% of
the pixel-to-pixel muscle T2 variations were shown to be due to imaging
noise. The results indicate that magnetic resonance imaging T2 cannot
reliably map active muscle on a pixel-by-pixel basis in normal subjects.
skeletal muscle; magnetic resonance imaging; recruitment; motor
units; spin-spin relaxation; muscle denervation
| |
INTRODUCTION |
IT IS WELL KNOWN that the
1H-NMR transverse relaxation time
(T2) of skeletal muscle water increases during exercise (11, 26).
Although the underlying cellular mechanism for the increase in muscle
T2 is not fully understood (3, 6, 13, 21), the phenomenon has been
exploited in many magnetic resonance (MR) imaging studies to map the
location (12, 14) and relative intensity (7, 11) of whole muscle
recruitment during various motor tasks. This practical application of
muscle "functional" MR imaging is justified by the empirical
observation that there is a good correlation between the
average T2 of a muscle or muscle group after exercise and other global
measures of recruitment intensity, such as surface electromyogram (1)
and work rate (15, 17).
Beyond mapping the recruitment of whole muscles, some studies indicate
that MR imaging can also be used to map regional variations in activity
within a muscle during exercise on a pixel-by-pixel level. This was
first suggested by Adams et al. (2), who observed distinct regional
variations in T2 within knee extensor muscles after fixed-rate
electrical muscle stimulation (EMS) at various voltages. In the study
of Adams et al., there was a good correlation between the number of
muscle image pixels showing increased T2 (nominally "active
muscle") and the force of knee extension. In contrast, the mean T2
of the active muscle was found to be independent of force.
The interpretation of these results was that muscle cells respond in an
on-or-off fashion during fixed-rate EMS, with recruitment order
determined by the pattern of current flow and the spatial distribution
of motor end plates within the muscle. Several recent studies (20, 22,
25) have extended the methods introduced by Adams et al. (2) to derive
an analogous estimate of "active muscle area" from MR imaging
data acquired after voluntary exercise. As in the original EMS study,
active muscle was defined on a pixel-by-pixel basis if a pixel's T2
exceeded a threshold T2 value, which was set at one standard deviation
(SD) above the mean T2 of pixels in the same muscle before the
exercise. Remarkably, these studies also found an excellent correlation
between "percent active area" of knee extensor muscles and the
intensity of voluntary exercise. Furthermore, resistance training (20)
and disuse atrophy (22) were accompanied by corresponding decreases and
increases in the percentage of total extensor muscle area that was
judged "active" during exercise at constant intensity.
Of course, single muscle cells are much smaller than the pixel
dimension achieved in a standard MR image. For example, with the
assumption of an image field of view (FOV) of 16 cm and a 256 × 256 pixel matrix, the pixel area is 0.4 mm2, i.e.,
(FOV/256)2. Even with the
assumption that the image slice is perpendicular to the long axis of
the muscle fibers, this pixel area must include 80-100 fibers, if
a mean fiber cross-sectional area of 4,000-5,000 µm2 is assumed (24). Therefore,
implicit to the idea that MR imaging can distinguish between active and
inactive areas within a muscle during exercise is the assumption that
there is a correlation between the location of fibers in a muscle and
their recruitment. This is a reasonable assumption for EMS, because in
that case, recruitment depends on the macroscopic flow of current from
the EMS electrodes. In contrast, recruitment during voluntary exercise depends on motor unit activation. Therefore, if active and inactive areas of muscle can be distinguished on a pixel-by-pixel basis by MR
imaging during voluntary exercise, then the muscle fibers from
different motor units must be nonrandomly distributed among the pixels,
such that individual pixels contain a preponderance of fibers from a
subset of the total motor unit pool. Although "clustering" of
motor unit fibers is a well-known feature of motoneuron disease (19)
and the location of motor unit territories is correlated with
recruitment order in some animal muscles (4), there is little anatomic
evidence for either of these fiber distribution patterns in muscles of
healthy human subjects (5).
The primary purpose of this study was to test the validity of the above
threshold method for estimating active muscle area by examining the
distribution of pixel T2 in muscles at rest and after moderate vs.
intense voluntary exercise. We reasoned that if the implicit assumption
of the threshold method is correct, then the pixel-to-pixel
distribution of T2 should be broadest after moderate exercise, ideally
approaching a bimodal distribution, because in that case some motor
units (and, therefore, some pixels) would be active while others would
not. At the opposite extreme, if the muscle cells of all motor units
were more randomly distributed across the muscle, then the T2 increase
would be spread across the muscle at all exercise intensities. In that
case, the pixel-by-pixel variation in T2 might largely reflect the
noise inherent in the T2 measurement and could not be used to reliably
map active vs. "inactive" pixels with a fixed threshold. The
experimental results are more similar to the latter extreme and,
moreover, are quantitatively consistent with a Monte Carlo simulation
that assumed a random distribution of motor unit territories across
muscles and a random distribution of motor unit cells within those
territories. However, the simulation does suggest that pixel T2
variation after exercise could be substantially greater in patients
with motoneuron disease.
| |
METHODS |
Subjects.
Seven adult subjects [5 men and 2 women, mean age 34 (range
20-47) yr] were recruited from the university community.
Subjects gave informed, written consent, and the study was approved by the University's Committee on Research Involving Human Subjects.
Exercise protocols.
Three single-limb weight-lifting exercises [biceps curl (5 subjects), ankle dorsiflexion (5 sujects), and knee extension (6 subjects)] were studied at two intensities: Max and
1/2Max. In each case, the Max exercise consisted of a single
repetitive bout of lifting a weight equal to 50% of the previously
determined one repetition maximum (1 RM) at a rate of 20 repetitions
per minute until failure (Table 1). The
1/2Max exercise consisted of a single bout with the same number
of repetitions performed at the same rate, but with the weight reduced
to 25% of 1 RM. Each subject first performed the Max exercise, and the
number of repetitions to failure was recorded. Approximately 30 min
later the subject performed the 1/2Max exercise using the
contralateral limb. In each case, two to three of the subjects used the
dominant limb for the 1-RM testing and Max exercise; the other subjects
used the nondominant limb.
MR imaging and analysis.
Immediately before and within 3 min after each exercise, the exercised
muscles were imaged on a 1.5-T GE Signa MR imaging machine (GE Medical
Systems, Milwaukee, WI). For leg muscles, axial fast spin-echo images
(256 × 128 matrix zero-filled to 256 × 256, repetition time = 2,000 ms, echo time = 30 and 60 ms, 9 or 10 1-cm-thick slices,
echo-train length = 4) were acquired via the body coil (knee extensors,
40 cm FOV) or via a standard 25-cm-diameter extremity coil (ankle
dorsiflexors, 16-20 cm FOV). For arm muscles, spin-echo images
(repetition time = 1,500, 16-cm FOV, other parameters as described
above) were acquired via the extremity coil. T2 images were computed on
a pixel-by-pixel basis from the two magnitude images with the
assumption of a single-exponential decay and after subtraction of the
mean magnitude of the background noise measured in a large region of
interest (ROI) outside the imaged limb. T2 images were analyzed by
tracing an ROI around the muscle group of interest with a
custom-written image analysis program. The T2 value (mean ± SD),
the total number of pixels in the ROI, and a histogram of the pixel T2
distribution were determined for the ROI in each slice. Results for the
whole muscle were computed by summing the results from the 9-10
images of each set. Active muscle area was estimated by the threshold
method, as described by Ploutz-Snyder et al. (22), except no correction was made for "nonmuscle pixels" within the ROI of the muscle at rest. Instead, visible nonmuscle areas, such as fat and vascular structures, were manually excluded during tracing of the ROIs.
Monte Carlo simulation of T2 variance.
The dependence of pixel T2 variance on motor unit size and cell
distribution was simulated using the model of muscle cell distribution
and recruitment developed by Fuglevand and Segal (16). The model
assumed a 200 × 200 grid of 50 × 50 µm square muscle
fibers (total 1-cm2 area, 4 × 104 cells) innervated by
100 motor units. Motor units ranged in size and number according to an
exponential function, such that there were many small units and few
large units (mean 400 fibers/unit, median 164 fibers/unit, smallest 14 fibers, largest 1,853 fibers). As in the original model (16), motor
unit territories were assumed to be square, with cell densities a
linear function of motor unit size ranging from 10 to 40 unit
fibers/mm2, resulting in a mean
territory area of 15.1 mm2. For
each trial, the territory of each motor unit was randomly assigned a
location on the grid, and the cells of this unit were randomly selected
from cells in the assigned territory (see Fig. 2 in Ref. 16). These
random assignments were made in order from the first (smallest) unit to
the 99th (second largest) unit. If a unit's territory exceeded the
boundaries of the grid, its territory was wrapped around to the
opposite side. The remaining cells were then assigned to the largest
unit, which was not recruited in the simulations. Recruitment was
assumed to be an "off-on" phenomenon in fixed order from the
smallest to the largest unit as a function of force (proportional to
total cells recruited) for four force levels: 10, 25, 50, and 75% of
maximum force. This corresponded to recruitment of 55, 73, 87, and 97 of the motor units, respectively. MR image pixels were modeled as a 20 × 20 pixel grid overlying the muscle grid; i.e., each pixel was
0.5 × 0.5 mm and contained 100 muscle fibers. Pixel T2
was assumed to be proportional to the fraction of "on" fibers in
the pixel and to range from 30 ms (no fibers on) to 40 ms (all fibers
on). Finally, random Gaussian noise (mean = 0, SD = 1, 2, or 3 ms) was
added to the pixel T2 values. The effect of this noise on the accuracy
of threshold classification of individual pixels as active vs. inactive
was computed by assuming a threshold at 1 SD greater than 30 ms (e.g., for noise SD = 2, the active threshold was 32 ms or 20% of the fibers
in a pixel active) and computing the percentage of correctly classified
pixels. Twenty-five trials of each of these "normal" muscle
simulations were computed, and the results were averaged.
The above normal simulation was modified to explore the effect of the
fiber clustering associated with partial muscle denervation due to
motoneuron disease. Fifty motor units were randomly selected for
elimination from the total population of 100 units assigned as
described above. The cells previously assigned to these units were
randomly reassigned to a motor unit innervating one of the eight
adjacent fibers or, if no such unit was available, to a random unit
innervating one of the 24 fibers within two squares. After this
reassignment of fibers, pixel T2 was recomputed for the four levels of
activity, again by averaging the results of 25 trials.
Statistics.
Data were analyzed using SPSS version 6.1 (SPSS, Chicago, IL). A
one-way ANOVA was used to test for significant differences between the
whole muscle ROI mean and the active muscle mean T2 values. Paired
t-tests with a Bonferroni adjustment
were used to test for significant differences between mean T2 values
across exercise intensities. The Brown-Forsyth procedure was used to test the homogeneity of whole muscle group T2 variance (18). All tests
were at the P < 0.05 level.
| |
RESULTS |
Experimental results.
Figure 1 shows representative axial images
of the three muscle groups after the Max exercises. There was an
obvious increase in intensity of the muscles expected to be recruited
after each of these exercises (i.e., biceps/brachialis, tibialis
anterior, and quadriceps for curl, ankle dorsiflexion, and knee
extension, respectively). As expected from previous studies, within
each recruited muscle group there was a roughly linear increase in mean
muscle T2 with increasing exercise intensity (Fig.
2, Table 2). In contrast to
previous studies (22), the mean T2 at "rest" was significantly
higher for the knee extensors and ankle dorsiflexors than for the elbow
flexors. Because spin-echo and fast spin-echo T2 pulse sequences are
subject to errors caused by imperfections in the refocusing pulses (8),
these differences in absolute T2 of the muscles before exercise may
arise in part from instrument errors. However, despite the differences
in initial T2, the changes in T2 with exercise were similar in the
three muscles, e.g., an 8- to 10-ms increase after the Max exercise.
View larger version (65K):
[in this window]
[in a new window]
|
Fig. 1.
Representative magnetic resonance (MR) images for biceps brachii (Bic,
A), tibialis anterior (TA,
B), and quadriceps (Quad,
C) muscle groups after a single
repetitive bout of lifting a weight equal to 50% of previously
determined 1 repetition maximum at a rate of 20 repetitions per minute
until failure (Max exercise).
|
|
View larger version (12K):
[in this window]
[in a new window]
|
Fig. 2.
Region of interest (ROI) transverse relaxation time (T2) values for
biceps brachii ( ), quadriceps ( ), and TA ( ) muscle groups at
each exercise condition. Values are means ± SE. 1/2Max, a
single bout with same number of repetitions as Max performed at same
rate as Max, but with weight reduced to 25% of 1 repetition maximum.
|
|
Figure 3 shows sample histograms of the
distribution of pixel T2 values summed across the whole muscles of
individuals at rest and after the two exercise intensities. Although in
each case the mean T2 increased with exercise, in no case was there a
distinct bimodal distribution. Thus there was no significant difference
in the pixel-to-pixel variance between the 1/2Max and the Max
exercise (Table 3). There were trends
toward increased variance after the exercises compared with rest,
although these trends were not significant according to the
Brown-Forsyth procedure.
View larger version (17K):
[in this window]
[in a new window]
|
Fig. 3.
Representative pixel T2 histograms for biceps brachii
(A), TA (B), and
quadriceps (C) muscle groups at rest (solid line) and
after 1/2Max (dotted line) and Max (dashed line) exercise.
|
|
Figure 4 shows the percent active muscle
area and the mean T2 of the active muscle, computed as described in
METHODS, for the three muscles. As
expected from previous studies (20, 22), the percent active muscle also
increased with exercise intensity. The mean T2 of the active muscle
also increased slightly with exercise intensity, but the magnitude of
this increase was small compared with the T2 change averaged across the
whole muscle.
View larger version (12K):
[in this window]
[in a new window]
|
Fig. 4.
Active muscle area (A) and active
muscle T2 (B) for biceps brachii
(), quadriceps (), and TA () muscle groups at each exercise
condition. Values are means ± SE.
|
|
The variation of the pixel-by-pixel T2 calculation depends on the
inherent signal-to-noise ratio (S/N) of the images, which in turn is a
complex function of the ratio of echo time to T2, the
receiver coil's properties, the pixel size and slice thickness, the
receiver bandwidth, and other factors. The contribution of this image
noise to the variation in pixel T2 can be estimated from images of a
homogeneous sample with similar T2 acquired at the same instrument
settings. For example, as shown in Fig. 5, image noise alone accounts for 1.43 ms (46%) of the SD of pixel T2 in
the resting anterior tibialis images of this study. Similar measurements showed that noise accounts for 1.0 ms (43%) and 1.6 ms
(39%) of the SD of pixel T2 in the resting biceps and quadriceps muscles, respectively.
View larger version (10K):
[in this window]
[in a new window]
|
Fig. 5.
Relationship between standard deviation of calculated pixel T2 and
reciprocal of signal-to-noise ratio (S/N) in images (echo time = 30 ms)
of a homogeneous CuSO4 solution
acquired at same instrument settings as TA images in this study. S/N of
solution was varied by varying image slice thickness (3-10 mm) and
repetition time (100-2,000 ms). T2 variation of solution (1.43 ms)
at same S/N level as muscle (mean 54:1) provides an estimate of muscle
T2 variation, which is due to imaging noise.
|
|
Simulation results.
Figure 6 illustrates the distribution
pattern of recruited fibers in single trials of the simulation of
normal and partially denervated muscle with 50% of the fibers active.
Figure 7 shows the summed pixel T2
histograms computed from 25 trials of these simulations at four fiber
activity levels. In the normal simulation, there is a modest increase
in T2 variance (Table 4) with activity compared with rest, but there is little change in variance between 25 and 75% activity and only slight divergence of the histogram shape
from a Gaussian distribution at the highest activity. On the other
hand, in the denervation model the variance of pixel T2 increases more
dramatically with exercise and depends more strongly on total fiber
activity level, and there is clear distortion of the histogram shape at
the higher activities.
View larger version (114K):
[in this window]
[in a new window]
|
Fig. 6.
Single trials of muscle simulation model at 50% total fiber activity
in normal (A) and denervated
(B) muscle. White areas, recruited
fibers; black areas, unrecruited fibers.
|
|
View larger version (20K):
[in this window]
[in a new window]
|
Fig. 7.
Computed distribution of pixel T2 values at rest and at 10, 25, 50, and
75% total fiber activity for 25 trials of normal simulation
(A) and denervation simulation
(B). Random Gaussian noise (mean = 0, SD = 1 ms) was added in each case.
|
|
View this table:
[in this window]
[in a new window]
|
Table 4.
Computed pixel T2 in simulations of normal and denervated muscle at
rest and at 4 activity levels, with Gaussian noise added
|
|
Finally, the simulation shows that the addition of Gaussian noise
profoundly degrades the accuracy of the threshold method for
individually classifying pixels as active vs. inactive. For example, at
the 25% fiber activity level, the percentage of correctly classified
pixels at noise SD = 1, 2, and 3 ms was 88, 65, and 59%, respectively.
Thus, with noise SD > 1 ms, the accuracy of the threshold method for
identifying individual active pixels is little better than a purely
random classification (i.e., 50% correctly classified). On the other
hand, the accuracy of threshold pixel classification is somewhat better
in the noisy denervation simulations (87, 71, and 65% for SD = 1, 2, and 3 ms, respectively), as would be expected from the greater
dispersion of pixel values during exercise in this model.
| |
DISCUSSION |
Pixel threshold mapping of active muscle by MR imaging.
Application of an absolute threshold criterion to map active pixels
within an MR image of muscle after voluntary exercise depends on two
implicit assumptions. First, the method assumes that the distribution
of motor unit cells among pixels is not random, such that, at
intermediate recruitment intensities, some pixels are substantially
more active than other pixels. Second, the method assumes that the
variance of the pixel T2 measurement is low compared with the expected
variance due to pixel recruitment. Our results indicate that neither of
these assumptions is correct for studies of normal subjects with
standard imaging techniques.
If the effect of pixel variance is ignored for a moment, the first of
the above assumptions suggests that, after moderate-intensity exercise,
when a muscle is not fully active, the pixel T2 histogram should
substantially broaden, ideally approaching a bimodal distribution. As
illustrated in Fig. 3, we found no evidence for such bimodal behavior
in the muscles of normal human subjects after voluntary exercise. On
the contrary, there was no significant difference in pixel-to-pixel
variance in any of the muscle groups after moderate exercise compared
with after exercise to exhaustion and, therefore, no evidence that
different pixels within the muscles behaved differently at these two
intensities. Although there was a trend toward increased variance after
exercise compared with rest, similar behavior was evident in the
simulation of normal muscle, which assumed random distributions of
motor unit cells among pixels, and of motor unit territories across the
muscle. Unfortunately, it is not possible to position a subject's limb
so precisely (i.e., within <1 pixel dimension) that the pixel
locations exactly correspond in images acquired before and after the
exercises and, thereby, demonstrate that the change in T2 is similar in
every pixel. Nonetheless, neither our data nor previously published T2
histograms (22) can exclude the hypothesis that the change in T2 occurs
throughout the recruited muscles in normal subjects at different
exercise intensities.
The application of an absolute threshold to map active pixels also
ignores the inherent variance of the pixel-by-pixel T2 calculation.
These variations in pixel T2 arise from two sources: 1) from the noise associated with
the imaging process and 2) from true
biological variation in the composition of the pixels. On the basis of
relationships similar to that shown in Fig. 5, it appears that in this
study the imaging noise per se contributed 1.0-1.6 ms to the total
variation in resting intramuscular T2. The results of the simulations
demonstrate that this noise can profoundly degrade the accuracy of the
threshold method. For example, at image S/N level one-half of that
achieved in this study (as might easily occur if slice thickness or FOV
were decreased to obtain increased spatial resolution), the threshold
classification of individual pixels as active or inactive after
moderate exercise would be little better than random, even with the
assumption that the T2 of the muscles at rest was perfectly homogeneous.
The remaining biological variation in pixel T2 must arise in part from
random inclusion of unresolved nonmuscle tissues (e.g., fat, small
blood vessels, and connective tissue) in the pixels. This inference is
supported by the observation that, in this study, imaging noise
accounted for the smallest fraction of the T2 variance in the
quadriceps muscle, which had the highest total T2 variance but the
lowest spatial resolution. In the presence of such variations in pixel
composition, application of a single absolute threshold criterion for
"activity" is clearly not appropriate, even in the absence of
imaging noise. This is nicely illustrated on a macroscopic scale by
comparing our results across the three muscles. Although the change in
T2 during the exercises was comparable in the three muscles, the
initial "resting" T2 was higher in the leg muscles than in the
arm muscle. Clearly, no single active threshold would be appropriate to
the three cases. Similarly, no single threshold can be applied to all
the pixels within a muscle, because there is also a broad distribution
of intramuscular pixel T2 at rest.
On the basis of the findings described above, we conclude that, given
the image resolution, S/N, and T2 variance presently achievable in
standard clinical MR imaging machines, the computation of active muscle
area on a pixel-by-pixel basis in normal subjects with use of a single
threshold T2 is not justified. Why then does the total active area
calculated by this method consistently correlate with exercise
intensity in this study (Fig. 4) and in previous studies (20, 22)?
Furthermore, why is it that the T2 of the "active area" is
relatively independent of exercise intensity, a result that seems
superficially consistent with the idea that active motor units fire at
a relatively constant rate once recruited (10), analogous to what
occurs during fixed-rate EMS? In fact, these two results are just
mathematical consequences of the threshold method for defining active
area in the presence of noise. First, with the assumption of a Gaussian
distribution of pixel T2 values at rest
where
f(x) is the number of pixels with T2 = x, µ is the mean T2, and 
is the SD, the active muscle
area is just
where
t is the threshold T2. As illustrated
in Fig.
8A, if the
mean of the distribution increases but the noise level is relatively
high, such that the SD does not change with exercise, then over a
considerable range the computed active area is just directly
proportional to the mean T2 of the distribution. Thus, in the presence
of noise, the observed correlation between total active area and
exercise intensity just reiterates the fact that the mean muscle T2
increases with exercise intensity and does not demonstrate that the
activity has been mapped to specific pixel locations. Second, the mean
T2 of the active muscle area, computed as described above, is
As
shown in Fig. 8B, this parameter does
not scale linearly with the overall mean T2, which can be easily
understood from the fact that the mean of the active area cannot be
less than the threshold T2. Thus the observation that the T2 of active
muscle is relatively invariant compared with the whole muscle T2 is
just a mathematical consequence of the threshold method for demarcating active muscle in noisy images and does not provide any additional information about the firing rate or other behavior of active motor
units.
View larger version (15K):
[in this window]
[in a new window]
|
Fig. 8.
A: theoretical plot of active muscle
area as a function of ROI mean T2. Resting ROI T2 = 27 ± 6 (SD) ms.
Each point is an increase in mean T2 by 0.25 SD; variance at each point
is constant. B: theoretical plot of
active muscle mean T2 as a function of ROI mean T2. Active muscle mean
T2 was based on active area derived in
A. Dashed line, threshold T2.
|
|
Justified applications of muscle functional MR imaging.
Although we observed no gross regional variations in T2 within muscles
in this study, our results do not exclude the possibility that
functional compartmentation of activity might be observed by MR imaging
in other muscles (9) or within these same muscles during different
types of exercises. On the contrary, our results confirm that the
muscle T2 averaged over whole muscles or fairly large ROIs (17)
increases linearly with exercise intensity (Fig. 2) and, therefore,
support the use of T2 as an index of muscle activity. In fact, as a
practical matter, the computation of total active muscle area as
performed by others (22) can also be used as an index of global muscle
activity, because, as shown above, this global index is roughly
proportional to mean muscle T2. However, we see no advantage to this
more complicated calculation, and the terminology, active area, is not
justified, because it falsely implies that activity has been localized
on a pixel-by-pixel level.
Despite the inability to resolve activity of single fibers or motor
units by MR imaging, comparison of the summed pixel T2 histograms with
those predicted by the Monte Carlo simulation suggests that MR imaging
could nevertheless yield information about the overall state of the
motor unit pool in human muscles. The simulation of normal muscle
predicts that the variance of pixel T2 should increase modestly after
exercise compared with rest but should be relatively constant across
intensities that recruit 25-75% of the total fibers (Fig. 7,
Table 4). This is remarkably similar to the observed behavior in our
healthy, relatively young subjects (Fig. 3). Of course, it cannot be
argued that the 1/2Max and Max exercises in this study
(performed at 25 and 50% of 1 RM) recruited 25 and 50% of the total
fiber populations, and it is probable that the number of recruited
units increased during the Max exercises as fatigue was approached.
Furthermore, the muscle simulation does not consider the effect of rate
modulation of force or the possibility that the T2 change in a muscle
fiber depends on unit firing rate or on fiber type. Nonetheless,
the agreement between the simulation and the results tends to confirm that the basic assumptions of the simulation are correct, e.g., that
the distribution of motor unit territories and of fibers within those
territories is random in healthy human muscle (16). In contrast, loss
of 
50% of the total motoneuron pool is common in patients with
motoneuron disease (19), and decreases up to 50% in motor unit number
have been measured in elderly subjects (23). In both cases, these
changes are known to be accompanied by increased motor unit size and
fiber clumping. The simulation of these processes indicates that the
muscle T2 distribution acquired after moderate exercise ought to be
substantially broadened in these patients compared with normal young
subjects and also that the accuracy of T2 threshold maps of activity in
these subjects would be greater. Therefore, if this simulation is
correct, then it may be possible to monitor the pathological changes
accompanying motor unit disease noninvasively by using a simple MR
imaging exercise test.
In summary, the results of this study show that the variance of MR
imaging pixel T2 in the recruited muscles of healthy subjects does not
substantially vary between different exercise intensities. Although MR
imaging does provide an empirical index of whole muscle recruitment and
may yield additional information on the overall state of the motor unit
pool, the use of MR imaging data to calculate active muscle area on a
pixel-by-pixel basis in normal subjects is not justified.
| |
ACKNOWLEDGEMENTS |
This work was supported by National Institute of Arthritis and
Musculoskeletal and Skin Diseases Grant AR-43903.
| |
FOOTNOTES |
The costs of publication of this
article were defrayed in part by the
payment of page charges. The article
must therefore be hereby marked
"advertisement"
in accordance with 18 U.S.C. §1734 solely to indicate this fact.
Address for reprint requests and other correspondence: R. A. Meyer,
Dept. of Physiology, Giltner Hall, Michigan State University, East
Lansing, MI 48824 (E-mail: ram{at}pslsun.psl.msu.edu).
Received 17 February 1998; accepted in final form 12 August 1999.
| |
REFERENCES |
1.
Adams, G. R.,
M. R. Duvoisin,
and
G. A. Dudley.
Magnetic resonance imaging and electromyography as indexes of muscle function.
J. Appl. Physiol.
73:
1578-1583,
1992[Abstract/Free Full Text].
2.
Adams, G. R.,
R. T. Harris,
D. Woodard,
and
G. A. Dudley.
Mapping of electrical muscle stimulation using MRI.
J. Appl. Physiol.
74:
532-537,
1993[Abstract/Free Full Text].
3.
Archer, B. T.,
J. L. Fleckenstein,
L. A. Bertocci,
R. G. Haller,
B. R. Barker,
R. W. Parkey,
and
R. M. Peshock.
Effect of perfusion on exercised muscle: MR imaging evaluation.
J. Magn. Reson. Imaging
2:
407-413,
1992[Medline].
4.
Armstrong, R. B.,
and
M. H. Laughlin.
Metabolic indicators of fibre recruitment in mammalian muscles during locomotion.
J. Exp. Biol.
115:
201-213,
1985[Abstract/Free Full Text].
5.
Basmajian, J. V.,
and
C. J. De Luca.
Muscles Alive: Their Functions Revealed by Electromyography. Baltimore, MD: Williams & Wilkins, 1985.
6.
Cole, W. C.,
A. D. LeBlanc,
and
S. G. Jhingran.
The origin of biexponential T2 relaxation in muscle water.
Magn. Reson. Med.
29:
19-24,
1993[Medline].
7.
Conley, M. S.,
R. A. Meyer,
J. J. Bloomberg,
D. L. Feeback,
and
G. A. Dudley.
Noninvasive analysis of human neck muscle function.
Spine
20:
2505-2512,
1995[Medline].
8.
Crawley, A. P.,
and
R. M. Henkelman.
Errors in T2 estimation using multislice multiple-echo imaging.
Magn. Reson. Med.
4:
34-47,
1987[Medline].
9.
English, A. W.,
S. L. Wolf,
and
R. L. Segal.
Compartmentalization of muscles and their motor nuclei: the partitioning hypothesis.
Phys. Ther.
73:
857-867,
1993[Abstract/Free Full Text].
10.
Erim, Z.,
C. J. De Luca,
K. Mineo,
and
T. Aoki.
Rank-ordered regulation of motor units.
Muscle Nerve
19:
563-573,
1996[Medline].
11.
Fisher, M. J.,
R. A. Meyer,
G. R. Adams,
J. M. Foley,
and
E. J. Potchen.
Direct relationship between proton T2 and exercise intensity in skeletal muscle MR images.
Invest. Radiol.
25:
480-485,
1990[Medline].
12.
Fleckenstein, J. L.,
L. A. Bertocci,
R. L. Nunnally,
R. W. Parkey,
and
R. M. Peshock.
Exercise-enhanced MR imaging of variations in forearm muscle anatomy and use: importance in MR spectroscopy.
AJR Am. J. Roentgenol.
153:
693-698,
1989[Abstract/Free Full Text].
13.
Fleckenstein, J. L.,
R. G. Haller,
S. F. Lewis,
B. T. Archer,
B. R. Barker,
J. Payne,
R. W. Parkey,
and
R. M. Peshock.
Absence of exercise-induced MRI enhancement of skeletal muscle in McArdle's disease.
J. Appl. Physiol.
71:
961-969,
1991[Abstract/Free Full Text].
14.
Fleckenstein, J. L.,
D. Watumull,
L. A. Bertocci,
R. W. Parkey,
and
R. M. Peshock.
Finger-specific flexor recruitment in humans: depiction by exercise-enhanced MRI.
J. Appl. Physiol.
72:
1974-1977,
1992[Abstract/Free Full Text].
15.
Fleckenstein, J. L.,
D. Watumull,
D. D. McIntire,
L. A. Bertocci,
D. P. Chason,
and
R. M. Peshock.
Muscle proton T2 relaxation times and work during repetitive maximal voluntary exercise.
J. Appl. Physiol.
74:
2855-2859,
1993[Abstract/Free Full Text].
16.
Fuglevand, A. J.,
and
S. S. Segal.
Simulation of motor unit recruitment and microvascular unit perfusion: spatial considerations.
J. Appl. Physiol.
83:
1223-1234,
1997[Abstract/Free Full Text].
17.
Jenner, G.,
J. M. Foley,
T. G. Cooper,
E. J. Potchen,
and
R. A. Meyer.
Changes in magnetic resonance images of muscle depend on exercise intensity and duration, not work.
J. Appl. Physiol.
76:
2119-2124,
1994[Abstract/Free Full Text].
18.
Keppel, G.
Design and Analysis: A Researcher's Handbook. Englewood Cliffs, NJ: Prentice Hall, 1991.
19.
McComas, A. J.,
V. Galea,
and
H. de Bruin.
Motor unit populations in healthy and diseased muscles.
Phys. Ther.
73:
868-877,
1993[Abstract/Free Full Text].
20.
Ploutz, L. L.,
P. A. Tesch,
R. L. Biro,
and
G. A. Dudley.
Effect of resistance training on muscle use during exercise.
J. Appl. Physiol.
76:
1675-1681,
1994[Abstract/Free Full Text].
21.
Ploutz-Snyder, L. L.,
S. Nyren,
T. G. Cooper,
E. J. Potchen,
and
R. A. Meyer.
Different effects of exercise and edema on T2 relaxation in skeletal muscle.
Magn. Reson. Med.
37:
676-682,
1997[Medline].
22.
Ploutz-Snyder, L. L.,
P. A. Tesch,
D. J. Crittenden,
and
G. A. Dudley.
Effect of unweighting on skeletal muscle use during exercise.
J. Appl. Physiol.
79:
168-175,
1995[Abstract/Free Full Text].
23.
Roos, M. R.,
C. L. Rice,
and
A. A. Vandervoort.
Age-related changes in motor unit function.
Muscle Nerve
20:
679-690,
1997[Medline].
24.
Saltin, B.,
and
P. D. Gollnick.
Skeletal muscle adaptability: significance for metabolism and performance.
In: Handbook of Physiology. Skeletal Muscle. Bethesda, MD: Am. Physiol. Soc., 1983, sect. 10, Chapt. 19, p. 555-631.
25.
Sloniger, M. A.,
K. J. Cureton,
B. M. Prior,
and
E. M. Evans.
Anaerobic capacity and muscle activation during horizontal and uphill running.
J. Appl. Physiol.
83:
262-269,
1997[Abstract/Free Full Text].
26.
Yue, G.,
A. L. Alexander,
D. H. Laidlaw,
A. F. Gmitro,
E. C. Unger,
and
R. M. Enoka.
Sensitivity of muscle proton spin-spin relaxation time as an index of muscle activation.
J. Appl. Physiol.
71:
84-92,
1994.
J APPL PHYSIOL 87(6):2107-2114
8570-7587/99 $5.00
Copyright © 1999 the American Physiological Society
TOP
ABSTRACT
INTRODUCTION
