Lipid content in the musculature of the lower leg assessed by fat selective MRI: Intra- and interindividual differences and correlation with anthropometric and metabolic data




To assess the muscular lipid content (LC) in different muscle groups of the lower leg by a magnetic resonance imaging technique working with chemical shift selective excitation, and comparison with anthropometric and metabolic data.

Materials and Methods

Examinations were performed in 67 volunteers (54 male/13 female, age 29 ± seven years) on a 1.5 T whole body imager, applying a highly selective spectral-spatial technique for fat selective MRI. LC was measured in six calf muscles and correlated with body mass index (BMI), percent body fat (PFAT), and insulin sensitivity (IS) of the subjects.


Mean muscular LC of all subjects was between 2.0% (Tibialis posterior [TP]) and 3.8% (Peroneus muscles) with female subjects showing a significantly higher LC in all muscle groups (P < 0.05 each). LCs correlated moderately with BMI (R between 0.39 [TP] and 0.53 [GM]) and with PFAT (R between 0.38 [TP] and 0.62 [GM]). Insulin-resistant subjects showed slightly but not significantly increased LC compared to insulin-sensitive subjects in BMI-matched subgroups.


The fat-selective MRI technique allows a reliable non-invasive measure of muscular lipids – even in muscle groups with inherent low LC – within a relatively short measurement time of about three minutes. The presented data reveal interesting interrelationships between LC and anthropometric and metabolic data, and therefore provide new insight into muscular fat metabolism. J. Magn. Reson. Imaging 2003;17:350–357. © 2003 Wiley-Liss, Inc.

RECENT STUDIES HAVE SHOWN a significant role of muscular lipids in the pathogenesis of insulin resistance and type 2 diabetes mellitus (1–7). Changes in the lipid metabolism occur even before clear changes in the blood glucose level can be ascertained (8, 9). Within former spectroscopic studies, mainly tibialis anterior muscle (TA) and soleus muscle (SOL) have been investigated, being of different muscle fiber type composition and functionality (10, 11). It has been shown that TA (predominantly glycolytic type II muscle fibers) has a clearly lower lipid content (LC) than SOL (mainly oxidative type I fibers). Furthermore, several spectroscopic studies have been performed concerning the regulation of intramyocellular lipids (IMCL) during exercise (12) and replenishment after workload (13). Short term interventions such as a fat-rich diet (14) or intravenous lipid infusion have also been performed (14, 15), showing rapid changes in IMCL content under distinct circumstances.

Volume selective proton magnetic resonance spectroscopy (1H MRS) is a relatively complicated and time-consuming method, requiring a lavish post-processing procedure to obtain the desired information. Quantification of the IMCL signal is furthermore aggravated by the overlapping signal of extramyocellular lipids (EMCL), dependent on the orientation of the muscle fibers in relation to the static magnetic field (16, 17). Furthermore, in single voxel techniques, only a region of a few cm3 of one specified muscle group is recorded, eluding a statement about the spatial distribution of the lipids. Besides 1H MRS, several magnetic resonance imaging (MRI) techniques, such as inversion recovery techniques (18–20), chemical shift selective imaging (21), phase-sensitive techniques (22, 23), or spectral spatial excitation (24–29), allow a separate depiction of water or fat component in any site of the human body. Superior selectivity of frequency-selective excitation compared to inversion recovery or saturation techniques have been described in the literature (21, 30), thus being better suited for visualization of low amounts of lipids in human musculature. Recently, a new technique with extraordinarily high sensitivity and specificity to the lipid pool has been proposed (31). Detection of lipid concentrations as low as 1% of the entire muscle signal is possible with this technique, allowing a quantification of lipids in muscle groups with inherent low LC.

So far, no data have been published showing the LC in different muscle groups of the lower leg and its correlation with anthropometric or metabolic data as percent body fat (PFAT), body mass index (BMI), and insulin sensitivity. The presented prospective study was performed applying a highly selective fat imaging technique (31) to obtain detailed information about the spatial lipid distribution, differences between the muscle groups, and interindividual variabilities. The aim was to determine typical patterns of the lipid distribution in the muscles, and possible insight into the regulation of lipids in several muscular and other compartments.

Anthropometric data were ascertained and insulin sensitivity of the subjects was determined by a euglycemic hyperinsulinemic glucose clamp in close temporal interval to the MR examinations to rule out short-term influences of the mentioned parameters.


Sixty-seven volunteers (54 male, age 28 ± four years; 13 female, age 29 ± eight years) participated in the prospective study. Although all volunteers were without medication and healthy (including normal blood glucose levels), it should be mentioned that they stem from a large cohort of metabolically characterized relatives of patients with Type 2 diabetes.

All subjects underwent examinations for determination of anthropometric data and insulin sensitivity as well as MR examinations for the assessment of the LC in different muscle groups of the lower leg. MRI and glucose clamp examinations were performed in the early morning. All participants were informed about the experimental procedures and gave written consent. The study protocol was approved by our local ethics committee.

Anthropometric Data and Glucose Clamp

Body mass index (BMI in kg/m2) and percent body fat (PFAT in percent, by body impedance measurement) were determined in all subjects. As the study cohort showed strong variations in BMI, subjects were divided into three groups: BMI1: 1) obese subjects (11/67) with a BMI > 27 kg/m2; 2) BMI2: normal weight subjects (28/67) with 23 kg/m2 < BMI < 27 kg/m2; and 3) BMI3: lean subjects (28/67) with BMI < 23 kg/m2.

For determination of the insulin sensitivity of the subjects, a euglycemic hyperinsulinemic glucose clamp was performed, with minor modifications of the DeFronzo protocol (32). For this procedure, plasma insulin is kept on a supraphysiological level of about 300–400 pM (hyperinsulinemia, HI) and plasma glucose is kept constant between 80 and 100 mg/dl (euglycemia). As HI stimulates peripheral glucose uptake, plasma glucose decreases, and glucose has to be infused to keep the euglycemic condition. The amount of glucose to be infused to maintain the euglycemic condition is used as a marker for insulin sensitivity. The glucose infusion rate (GIR in μmol · kg–1 · min–1) was calculated.

Subjects were instructed to maintain a standard diet and to refrain from heavy exercise for at least three days before the metabolic tests. The study population was further subdivided into two groups, classified as insulin resistant (IR; GIR < 40.0 μmol · kg–1 · min–1), or insulin sensitive (IS; GIR ≥ 40.0 μmol · kg–1 · min–1).

Anthropometric and metabolic data are listed in Table 1.

Table 1. Anthropometric and Metabolic Classification of the Study Population
  1. BMI = body mass index in kg/m2, PFAT = percent body fat, GIR = glucose infusion rate in μmol · kg−1 · min−1.

>2711 (8m/3f)29.3 ± 5.830.6 ± 3.329.3 ± 8.325.8 ± 7.5
23–2728 (24m/4f)30.4 ± 6.824.8 ± 1.321.7 ± 5.637.6 ± 12.6
<2328 (22m/6f)27.9 ± 6.621.1 ± 1.316.3 ± 4.843.0 ± 11.4

Magnetic Resonance Examinations

MR examinations were performed on a 1.5 T whole body unit (Vision, Siemens, Erlangen, Germany). The circularly polarized extremity coil of the manufacturer was used as combined transmitter/receiver coil. Subjects were in supine position with the most extended part of the right calf in the center of the coil.

T1-Weighted SE-Imaging

Axial T1-weighted spin-echo images were recorded for depiction of the individual muscle groups (measurement parameters: echo time TE = 12 msec, repetition time TR = 650 msec, matrix 256 × 256, slice thickness 10 mm, acquisition time TA = 1:08 minutes). Selection of the regions of interest (ROIs) within the muscle groups for evaluation of the LC in the corresponding fat-selective image was performed in the T1-weighted image, since the borders of the muscle groups are more clearly visualized in those images. Figure 1 shows an axial T1-weighted image of a 34-year-old male subject with the relevant ROIs drawn in six muscle groups. For evaluation, the tibialis anterior (TA), tibialis posterior (TP), soleus (SOL), gastrocnemius lateralis (GL) and medialis (GM), and peroneus longus et brevi (PLB) were selected. Extensor digitorum longus (EDL) was not evaluated. Regions including subcutaneous fat or compact bone of the tibia or fibula were carefully excluded from the ROIs. Fatty septa between the muscle groups were also excluded, whereas fatty tissue between the muscle fiber bundles (EMCL) was covered by the ROIs. In addition to the ROIs in the muscle groups, two circular ROIs were selected in the fat selective images: one in the tibial bone marrow (BM) for determination of signal intensity of pure fat, and the other in an object-free part of the image for determination of the noise level (N).

Figure 1.

Axial T1-weighted spin-echo image of the calf of a 34-year-old male volunteer. Regions of interest (ROI) for evaluation of the muscular lipid content are drawn in tibialis anterior (TA), tibialis posterior (TP), soleus (SOL), gastrocnemius medialis (GM) and lateralis (GL), and peroneus longus et brevis (PLB). Additional ROIs in tibial bone marrow (BM) for determination of fat reference, and in an object free part of the image for determination of noise level (N, highlighted).

Fat-Selective Imaging

A recently developed gradient echo imaging technique was applied, with highly selective excitation of a Larmor frequency range corresponding with methylene and methyl protons of fatty acids (31). Repetition time TR was set to 50 msec, echo time TE was 16 msec. Spectral-spatial excitation was performed with six equidistant pulses with nearly binomial amplitude ratios. Considering the relaxation characteristics of methylene protons (T1 = 300 msec and T2 = 90 msec), and regarding Ernst's angle for optimal signal yield under these circumstances, flip angles of the six excitation pulses (Hanning-filtered sinc pulses) were set to 2° – (–8°) – 14° – (–14°) – 8° – (–2°) resulting in a total excitation flip angle of 48° for lipids, but 0° for water. The period of the single pulses was set to 2.32 msec to obtain a frequency difference of 215 Hz between maxima and minima of excitation, according to the chemical shift between water and methylene protons at 1.5 Tesla. Transverse magnetization was spoiled after signal recording. Readout time was set to 12.8 msec, resulting in a relatively low receiver bandwidth of 78 Hz per pixel and a low noise level. Field of view was chosen at 180 mm with a matrix size of 204 × 256 (pixel size 0.88 × 0.70 mm). To obtain a sufficient signal to noise ratio, slice thickness was set to 10 mm, and 20 acquisitions were recorded in a total measurement time of three minutes, 25 seconds.

Analysis of Muscular Lipid Content

LCs of the six calf muscle groups were determined from the fat selective images using a suitable self-made post-processing procedure (Matlab 6.1, The MathsWorks Inc.). Irregularly shaped ROIs were drawn in the T1 weighted SE-images regarding the geometry of each muscle group (Fig. 1) and then put on the fat selective images as described above. Signal ratios between the muscle groups and tibial bone marrow were calculated. Mean values were corrected for noise (33). Percentual muscular LC is

equation image

Statistical Analysis

For each muscle group, mean values and standard deviations (SD) of LC were calculated. Statistical analysis was performed using SigmaStat software tools (Jandel Scientific). Linear regression analyses were performed to evaluate the correlations between the LC of the different muscle groups, and between LC and anthropometric or metabolic data. Significance of differences between the data were analyzed using a two-sided Student's t-test. A P value of less than 0.05 was considered as statistically significant.


All volunteers were able to undergo the mentioned examinations. The fat-selective images clearly revealed different patterns of spatial lipid distribution, as visualized in Figure 2. The image in Figure 2a stems from a 32-year-old IS male volunteer with very low LC in all muscles, except some minor fatty septa. In contrast, a 34-year-old IR male volunteer shows clearly higher LC in most muscles, combined with a homogenous background signal in SOL, GM, and GL in Figure 2b. Figure 2c and 2d provide other examples with marked differences in the lipid distribution: The 27-year-old volunteer in Figure 2c is characterized by a low LC in TA and TP, and a considerably higher LC in the other muscles with a very intense background signal, whereas the 34-year-old female volunteer in Figure 2d has predominantly fatty septa in all muscles, with a clearly lower background signal.

Figure 2.

Fat-selective images of the calf, recorded using the spectral-spatial excitation sequence proposed in the text. a: 32-year-old male IS subject. b: 34-year-old male IR subject. c: 27-year-old male IR subject with high lipid content ratio SOL/TA. d: 34-year-old IS female subject.

The measured LC values of the muscle groups were analyzed under different points of view: intra- and inter-individual variations of LC of the muscle groups were assessed in all participants. Gender-specific differences and correlations of muscular LC values with BMI, PFAT, and GIR were investigated based on suitable subgroups of our cohort.

Interindividual Differences and Correlations

Mean LC values were clearly different for the muscle groups examined. Lowest LC was found in the tibialis muscles (LCTA = 1.86 ± 1.14%, LCTP = 2.01 ± 0.87), followed by the two gastrocnemius muscles, both showing a very similar LC (LCGL = 2.40 ± 1.27, LCGM = 2.41 ± 1.34). LC was clearly higher in SOL with LC = 2.93 ± 1.32%, and highest in PLB (3.78 ± 2.15%). Mean values, SDs, and ranges of the whole study group are given in the upper column of Table 2 and visualized in Figure 3. Except for the differences between TA/TP and GL/GM, all differences in LC of the muscle groups were statistically significant (P < 0.05).

Table 2. Muscular Lipid Content (LC) of the Different Calf Muscles in Dependence on BMI
  • a

    P < 0.05 between BMI1 and BMI2.

  • b

    P < 0.05 between BMI1 and BMI3.

  • c

    P < 0.05 between BMI2 and BMI3.

  • BMI = body mass index in kg/m2, TA = tibialis anterior muscle, TP = tibialis posterior muscle, SOL = soleus muscle, GM = gastrocnemius medialis muscle, GL = gastrocnemius lateralis muscle, PLB = peroneus longus et brevi muscles.

All1.9 ± 1.12.0 ± 0.92.9 ± 1.32.4 ± 1.32.4 ± 1.33.8 ± 2.2
>272.8 ± 1.6ab2.3 ± 0.93.8 ± 2.0b3.5 ± 2.3b3.3 ± 2.1ab4.9 ± 3.4b
23–271.7 ± 0.9a2.2 ± 0.9c3.0 ± 1.22.5 ± 0.9c2.3 ± 0.9a4.0 ± 2.0c
<231.6 ± 0.9b1.7 ± 0.8c2.5 ± 1.0b1.9 ± 1.0bc2.1 ± 1.0b3.1 ± 1.4bc
Figure 3.

Muscular lipid content (LC) of the six different calf muscles. Data are given as mean and SD.

Muscular LC showed high interindividual variability in all muscle groups. The variation of individual data was 0.7% – 4.3% in TP (smallest range), and 1.1% – 12.2% in PLP (widest range), as expressed by the bars in Figure 3.

Furthermore, the individual ratios of LC values between two different muscle groups of some individuals differ strongly. The signal ratio between SOL and TA is in a range between 0.63 and 5.86, with some of the volunteers (10/67) having even a higher LC in TA than in SOL. This difference is visualized in the fat selective images in Figure 2c and 2d.

Analyzing the LC values in the three BMI-groups, a clear tendency of higher muscular fat in all muscles for obese subjects was evident. Mean LC was always highest in group 1 (BMI > 27 kg/m2) and lowest in group 3 (BMI < 23 kg/m2). Data are comprised in Table 2. Lowest BMI-related differences resulted in TA between group 2 and group 3, and in TP between group 1 and group 2. Group 2 and group 3 differed significantly in TP (P = 0.02).

To figure out common regulation mechanisms or common fiber type compositions of the different muscle groups, correlation coefficients for LC of all subjects were determined by linear regression analysis. Values are given in Table 3. The highest correlation resulted between the medial and the lateral gastrocnemius muscles (rGM/GL = 0.90). In contrast, LC of TP is only slightly correlated to the LC of all other muscles, with the lowest correlation to TA (rTA/TP = 0.53). Correlation coefficients between GM and GL and other muscle groups X were relatively high (rGM/X = 0.55–0.90 and rGL/TP = 0.53–0.90).

Table 3. Correlation Coefficients of LC of Different Calf Muscles
  1. TA = tibialis anterior muscle, TP = tibialis posterior muscle, SOL = soleus muscle, GM = gastrocnemius medialis muscle, GL = gastrocnemius lateralis muscle, PLB = peroneus longus et brevi muscles.

TP 0.590.530.550.71
SOL  0.800.820.77
GL   0.900.81
GM    0.76

Gender-related differences were assessed in subgroups of nine male (age 27.0 ± 3.1 years) and nine female (age 26.9 ± 6.3 years) volunteers, matched for BMI (22.8 ± 2.1 kg/m2 vs. 22.4 ± 2.9 kg/m2) and GIR (32.7 ± 10.1 μmol · kg−1 · min−1 · pM−1 vs. 32.6 ± 10.7 μmol · kg−1 · min−1 · pM−1). Female volunteers showed significantly higher LC in all muscles, as depicted in Figure 4. In contrast to males with low LC in TA, the female subgroup showed LC in TA higher than in all other muscle groups, except PLB.

Figure 4.

Gender-specific lipid content (LC) of the evaluated calf muscles. Left bars: male subjects; right bars: female subjects. ◊ = P < 0.05; ‡ = P < 0.01; * = P < 0.001.

Correlations With Anthropometric and Metabolic Data

Regarding the entire cohort, muscular LC showed clear correlations with BMI and PFAT. LC values of TA, GL, and GM were stronger correlated to PFAT than to BMI. Correlation coefficients are given in Table 4. The data show lowest correlation coefficients for TP.

Table 4. Correlation Coefficients Between LC and BMI, PFAT, and GIR of All Volunteers
  1. BMI = body mass index in kg/m2, PFAT = percent body fat, TA = tibialis anterior muscle, TP = tibialis posterior muscle, SOL = soleus muscle, GM = gastrocnemius medialis muscle, GL = gastrocnemius lateralis muscle, PLB = peroneus longus et brevi muscles.


Including all volunteers, LC and GIR were negatively correlated in all muscle groups, as listed in Table 4. In principle, this could be mainly an effect of BMI and/or PFAT, which are often higher for insulin-resistant subjects. For this reason, relations between LC and GIR were determined within the BMI-groups mentioned above. Within group 1 (obese), all volunteers were classified as IR, eluding a comparison between IS and IR. In group 2 (normal weight), IR subjects (19/28) showed a higher LC in all muscles compared to the IS subjects (9/28). However, the differences were statistically not significant. In group 3 (lean) the LC were very similar between IR (8/28) and IS (20/28) subjects, except TA, just missing statistical significance (P = 0.06). Figure 5 depicts the mean LC and the SD for the subgroups mentioned above.

Figure 5.

Relationship between lipid content (LC) and insulin sensitivity. Data are given as mean and SD. Subjects are subdivided in three BMI-groups: 1) overweight (BMI > 27 kg/m2); 2) normal weight (BMI between 23 and 27 kg/m2); and 3) underweight (BMI < 23 kg/m2). Furthermore, LC of IR and IS subjects are shown separately except for group 1, where all subjects were classified as IR.


Spatially resolved assessment of the LC in different calf muscles was performed in a cohort of 67 healthy volunteers. A reliable measure of LC was possible within a relatively short measuring time of about three minutes, and with a straightforward and almost user-independent post-processing procedure. The results showed significant differences in LC of the examined muscle groups, which could be partly explained by the known differences in their muscle fiber distribution. Muscles with predominantly fast twitch type 2 fibers (TA, TP, GL, GM) are characterized by a clearly lower LC than muscles with predominantly slow twitch type 1 fibers (soleus, peroneus) (10, 11). This is in good agreement with spectroscopic data evaluated by Hwang et al (34), who quantified regional differences in intramyocellular (IMCL) and extramyocellular lipids (EMCL) in tibialis anterior (TA), tibialis posterior (TP), and soleus muscle (SOL).

Correlation between the LC values of different muscle groups by linear regression revealed strong intermuscular relationships mainly for the GL, GM, SOL, and PLB muscles, whereas the LC of the two tibialis muscles was less correlated to the LC of the other muscles. A possible explanation for this finding is the differing amount of macroscopic fatty septa between fiber bundles in TA and TP, which might be under the control of other genes and external factors than the other lipid compartments examined. The role of these small fatty septa and influencing factors are as yet unknown. In TP, LC shows the lowest variability and is not markedly higher even in obese subjects. In TA, it has to be taken into consideration that, in some cases, parts of the extensor digitorum longus muscle (EDL), which lies next to TA (see Fig. 1), are included in the selected ROI of TA. Unambigous separation of these two muscles is not always possible in T1-weighted images due to the lack of macroscopic fatty septa or fasciae.

It should be mentioned that LC measurements by the applied imaging approach might be slightly influenced by unavoidable artificial signal contributions caused by false spatial encoding in muscle groups located close to subcutaneous fat or bone marrow, as discussed in (31). LC values might be overestimated in those cases.

Possible influencing factors on the muscular LC are the personal training status of the subjects and variations in muscle fiber type distribution, as a single muscle consists of a mixture of all muscle fiber types and fiber type shares can show marked inter-individual differences (35). However, as neither the training status of the volunteers (VO2max) nor muscle fiber type composition has been determined, the influences of these factors remain speculative in this study.

Significant gender-related differences in LC were observed in subgroups of nine male and nine female subjects matched for age, BMI, and insulin sensitivity. Female subjects showed significantly higher LC in all muscles, which was most pronounced in TA. Perseghin et al (36) reported on a higher IMCL in female TA compared to the males, whereas they did not establish differences in SOL muscle. This discrepancy is explainable, as only the IMCL content of a small volume of interest within the muscle was included in the paper of Perseghin, whereas the results shown in our study are based on the sum of both portions, IMCL and EMCL, over the entire muscle cross-section.

Correlation of muscular LC with anthropometric data revealed a slightly stronger relation between LC and percentual body fat (PFAT) than between LC and body mass index in the whole cohort. Stein et al (4) and Forouhi et al (37) have already showed a correlation between intramyocellular lipids (IMCL) and body mass index (BMI) by single voxel spectroscopic studies, whereas Hwang et al did not find a significant correlation in TA and TP for IMCL or EMCL alone, but in all muscles (TA, TP, and SOL) for the combination of both (34).

In contrast to previous studies focusing on the relationship between IMCL and insulin sensitivity (4–7), only slight differences between insulin-sensitive and insulin-resistant subjects have been shown in our BMI-matched subgroups. Highest differences within the BMI-subgroups resulted in the TA of normal weighted volunteers (P = 0.06), whereas lean subjects had almost unique LC (P = 0.97). The imaging technique in our study provides relatively high spatial distribution, but does not allow differentiation between IMCL and EMCL, as spectroscopy did in the former studies. This might be the main drawback of the technique, as mainly the active IMCL compartment is thought to be involved in the pathogenesis of skeletal muscle insulin resistance and type 2 diabetes mellitus.

The presented study reveals pronounced intra- and interindividual differences between the muscular fat content in several muscle groups. Examinations of the reaction of muscular lipids on external factors such as training or fasting, and analysis of lipid distribution patterns dependent on genetic polymorphisms are further steps to a more complete knowledge on the interaction of muscular fat and metabolic pathways.


The authors gratefully thank the members of Siemens Medizintechnik (Siemens AG, Erlangen, Germany) for technical assistance.