Standardized assessment of whole body adipose tissue topography by MRI

BESIDES GENERAL ANTHROPOMETRIC and metabolic measures, the amount of fat as well as its regional distribution is considered to play an essential role in the pathogenesis of insulin resistance and type 2 diabetes mellitus (1-3). Common techniques for quantification of body fat mass as body impedance analysis (BIA) or underwater weighing (4) only provide information about the total amount of adipose tissue, whereas a detailed analysis of different lipid components is not available. Other anthropometric methods for the quantitative assessment of local fat—such as measurement of skin-fold thickness, waist circumference, or waist-to-hip ratio (WHR)—provide a simple and useful estimation of the proportion of abdominal fat (5), but are also not able to precisely distinguish between visceral adipose tissue (VAT) and subcutaneous abdominal adipose tissue (SCAT). There is evidence that mainly VAT, which is morphologically and functionally different from SCAT, is associated with insulin resistance, hyperinsulinemia, dyslipidemia, and hypertension (6-9).

Imaging techniques, particularly computed tomography (CT) and magnetic resonance imaging (MRI) allow to distinguish between VAT and SCAT (10) in each recorded slice. Accuracy of both modalities has been validated against dissection in human cadavers (11, 12). As CT is often applied as a single slice modality (13) due to the inherent drawback of exposure to ionizing radiation for the volunteers, MRI seems to be the most promising imaging tool for assessment of total body adipose tissue distribution (14, 15). This holds especially for follow-up studies, which are of increasing interest in research projects concerning metabolic diseases.

In the past, several MRI techniques have been proposed for measuring body fat composition and for quantification of the mentioned fat compartments. T1-weighted imaging by spin-echo or gradient-echo techniques were applied (14-16), in which fat appears bright due to its inherent shorter longitudinal relaxation time (T1 of fat is approximately 300 msec vs. about 1000 msec for most water-containing tissues (17)). Chemical shift selective (CHESS) imaging (18-20), or DIXON-techniques (21, 22) were also proposed, but all of the mentioned techniques have inherent advantages and drawbacks (21).

Several cross-sectional (16, 23-25) and longitudinal studies (26-28) have impressively demonstrated the unique potential of MRI for determination and quantification of the different fat compartments of the human body. However, up to now, mainly the amount of adipose tissue was analyzed without taking into account the topography of its distribution.

The aim of the present study was to perform a detailed analysis of whole body topography of adipose tissue in a cohort of subjects at risk for type 2 diabetes mellitus. A strategy for standardization of adipose tissue profiles from head to feet is presented, which enables a direct comparison of subjects with different body structure and large variations in anthropometric parameters. After standardization, several well-defined subgroups of our cohort were compared in order to figure out characteristic differences in their body fat distribution.

MATERIALS AND METHODS

A total of 80 healthy nondiabetic volunteers (40 females, 40 males) underwent MR whole-body imaging for determination of fat distribution. In addition, usual anthropometric data were assessed from all participants and metabolic examinations for determination of insulin sensitivity were performed. All volunteers were recruited within the framework of the Tuebingen Lifestyle Intervention Program (TULIP) study, which is designed as a longitudinal intervention study in order to examine people at risk to develop type 2 diabetes. Inclusion criteria for participation are body mass index (BMI) > 27 kg/m² (obese or overweight), family history of type 2 diabetes, impaired glucose tolerance, and/or history of gestational diabetes in females. Participants were informed in detail about the experimental procedures and gave written consent. The study protocol was approved by our local ethics committee.

T1-Weighted MRI of the Whole Body

MR examinations were performed in the early morning after an overnight fasting period on a 1.5-T whole body imager (Magnetom Sonata; Siemens Medical Solutions, Erlangen, Germany). For determination of whole-body fat distribution, an axial T1-weighted fast spin echo technique with an echo train length of 7 was applied. Measurement parameters: TE/TR 12 msec/490 msec, slice thickness 10 mm, 5 slices per sequence, 10 mm gap between the slices. Field of view was 450 mm to 530 mm depending on the extension of the volunteer. A 256 × 178 matrix was recorded in a measuring time of 12 seconds, allowing breathhold examinations in abdominal regions. Table shift was set to 10 cm. Volunteers were in prone position with the arms extended and data were collected from fingers to toes. The body coil was used as combined transmit/receive coil. Total examination time was between 20 and 25 minutes including one rearrangement, as total table feed of the MR-imager is limited to 110 cm. In order to guarantee identical slice positions after repositioning, volunteers were marked at the iliac crest.

Complete reproducibility measurements were performed in three volunteers. Both examinations in each volunteer were performed on the same day after new positioning.

In total, 100 to 130 images were obtained depending on the size of the volunteer. Figure 1 schematically shows the positioning of the volunteers and exemplary images of different body regions of a 39-year-old male volunteer.

Postprocessing of the images was performed on a personal computer applying a home-written segmentation program based on Matlab (Mathworks, Inc.). For this purpose two threshold values were set, the first for determination of the noise level in object free parts of the image, the second for differentiation between lean tissue and fat. The threshold value for separation of lean and fat tissue was automatically set to the nadir of bright pixels corresponding to adipose tissue. Its value could be slightly varied manually by visual inspection of the image in order to correct for smaller inconsistencies, as might arise in regions with inhomogenous signal illumination (see Fig. 1). Besides subcutaneous fat, bright fatty bone marrow of the extremities also contributes to adipose tissue and is included in the analysis.

Tissue volumes were calculated by multiplying the corresponding number of segmented pixels by the in-plane pixel dimensions and the slice thickness: total tissue volume (TT) including all pixels with signal intensities above the noise level, adipose tissue volume (AT) including all pixels above the second threshold value, visceral adipose tissue (VAT) and abdominal subcutaneous adipose tissue (SCAT) by manually drawing two regions of interest. Figure 2 depicts exemplary images recorded on the level of the umbilicus of a female (Fig. 2a) and a male volunteer (Fig. 2b), showing the segmentation in the original images (I) for determination of SCAT (II) and VAT (III).

As the interslice gap corresponds to the slice thickness, volumes between adjacent slices are calculated by simply doubling the volumes of the slices.

In order to standardize the profiles of volunteers with different body size (ranging between 154 and 193 cm in our cohort) and bodily structure, each individual dataset was divided in three parts and interpolated to a defined number of sampling points:

• 1

  Lower extremities (LE), ranging from the heel bones to the head of the femur. This body part originally contained 45–54 recorded slices and was interpolated to 70 sampling points for each volunteer.

• 2

  Trunk (T), from head of femur to the head of humerus. This body part originally contained 26–41 recorded slices and was interpolated to 50 sampling points for each volunteer.

• 3

  Head and upper extremities (UE), from the head of humerus to the wrist. This body part originally contained 25–35 slices and was interpolated to 40 sampling points for each volunteer.

The principle is shown in Fig. 3. Data were linearly interpolated using a standard Matlab routine. The error caused by this interpolation algorithm regarding the integrated volumes of TT and AT of the entire body is lower 0.5%.

RESULTS

Univariate linear correlation coefficients between MR-derived parameters of total tissue (TT), % adipose tissue (%AT), visceral and abdominal subcutaneous adipose tissue (VAT and SCAT), and anthropometric data, as well as insulin sensitivity (ISI) are given in Table 2 for females and males separately. The correlation between TT and weight was r = 0.99 for both females and males, indicating good reliability of the method. Correlation coefficients between MR-derived parameters and anthropometric measures were generally higher for females compared to males, whereas correlations with ISI were very similar. Female volunteers showed a significantly lower amount of VAT but higher abdominal SCAT and AT compared to males. This relation is quantitatively depicted in Fig. 4a for AT and VAT, and in Fig. 4b for AT and abdominal SCAT.

The whole-body topography of evaluated tissue compartments is presented in Fig. 5 for all participating female volunteers. In Fig. 5a, absolute values (mean/SEM) of total tissue (TT, closed circles), adipose tissue (AT, squares) and visceral adipose tissue profiles (VAT, triangles) are demonstrated. Figure 5b shows the percentage of AT and VAT. Body contours can clearly be assigned (compare Fig. 5c).

Comparison of Age Groups

There were no consistent differences in AT profiles between the selected age groups for both females and males (profiles not shown). However, in subgroups matched for BMI (N = 4, 29 kg/m² < BMI < 31 kg/m²) and PFAT (21–29%), a clearly increasing amount of VAT resulted for increasing age, as demonstrated in Fig. 6.

Comparison of BMI Groups

In all BMI groups, females are characterized by a significantly higher %AT compared to males. Figure 7 exemplarily depicts standardized %AT profiles of BMI₂ (30 kg/m² < BMI < 35 kg/m²) showing this effect in all body regions (Fig. 7a: lower extremities, LE; Fig. 7b: trunk, T; Fig 7c: upper extremities, UE). Differences are most pronounced in lower leg and thigh (Fig. 7a) and clearly smaller in the trunk (Fig. 7b). Furthermore, the lower amount of VAT in females is figured out, as females are characterized by clearly lower %VAT in all sampling points of the profiles (between 7% and 55% of the mean values), but higher total %AT.

Comparison of Groups with Different Insulin Sensitivity

Two subgroups with five subjects, each characterized by different insulin sensitivity but matched anthropometric data (age, BMI, PFAT) were selected from each gender group, and %AT profiles were compared. Figure 8 shows the profiles of the insulin sensitive (IS, □) and insulin resistant (IR, ▪) males. The AT distribution is very similar in the lower extremities (Fig. 8a). In the trunk, local differences in AT distribution become visible as indicated by the arrows in Fig. 8b: IR subjects are characterized by slightly higher VAT, and slightly higher AT in the shoulder/neck region compared to the IS subjects. On the other hand, IR subjects seem to be characterized by lower AT share in upper extremities (Fig. 8c). The reported differences were also observed for IR and IS females (profiles not shown).

DISCUSSION

Quantitative analysis of adipose tissue is of particular interest due to the public health problem with increasing prevalence rates of obesity and associated metabolic diseases as type 2 diabetes mellitus and cardiovascular risk. It has been shown that mainly visceral adipose tissue seems to be involved in the pathogenesis of insulin resistance and type 2 diabetes (1-3). In contrast, other authors reported that subcutaneous rather than visceral adipose tissue volume have had a stronger correlation with insulin resistance (30, 31). So there is still debate regarding which of these fat depots is more important in determining insulin resistance. However, looking not only at the amount of these adipose tissue compartments but also on their topography might bring new and important information which was not considered up to now.

It has been demonstrated in earlier work (e.g., 23–28) that magnetic resonance imaging offers a unique, noninvasive tool for a reliable, noninvasive determination of fat distribution in all body regions. Some recently performed studies were restricted to a defined body region, a defined number of slices independent on body size (28), or even single slice contemplations (32). Other approaches examined the entire body of the subjects but without taking into account the topography of adipose tissue (14, 16, 24, 27). However, the problem with different body structures in interindividual comparison of subjects remained. The present approach provides a simple and time-saving strategy for assessment and standardization of individual adipose tissue distribution in the entire body. A detailed analysis of adipose tissue distribution is possible, and therefore—due to standardization using defined internal markers—a reliable comparison of subjects with different body structure. After standardization, interesting groups of subjects can be analyzed for differences in their regional adipose tissue distribution. The presented studies indicate that groups of insulin-sensitive and insulin-resistant subjects that are matched for anthropometric data did not only confirm the earlier reported differences in VAT, but also show differences in the shoulder/neck region and—with opposite sign—in the upper extremities. It has to be mentioned that these results are preliminary due to the low number of included volunteers and have to be confirmed in a larger population.

Standardized recording of adipose tissue profiles will serve as an important tool in the analysis of effects resulting from lifestyle interventions such as diet or training. Not only the reduction but also the redistribution of adipose tissue in different body regions can be assessed. This might especially be interesting in subjects that are characterized by improved metabolic parameters without weight loss or reduced PFAT, as already described for medical interventions (33). The proposed technique might also be helpful for the search for individual predictive factors for potential metabolic diseases.

Some limitations of the proposed approach have to be taken into consideration. First, the gaps between the recorded slices could lead to minor errors by the interpolation of the data. Second, postprocessing of the images is not yet fully automated and therefore–besides the manpower requirement—minor inaccuracies could result from setting threshold values for adipose tissue and manual definition of subcutaneous or visceral adipose tissue in abdominal images. Third, signal of intestinal contents with short T1 could be interpreted as visceral adipose tissue. In our cohort, examinations were performed after an overnight fasting period and therefore these contaminations were avoided.

Besides imaging modalities, spectroscopic examinations have been performed showing a close association between insulin resistance and fat deposition in mainly water containing tissues including skeletal muscle (intramyocellular lipids, IMCL) (34-36) or liver (hepatic lipids, IHL) (37, 38). However, as spectroscopy is restricted to a few MR centers, whereas standard imaging techniques are available on each MR unit, imaging studies can be performed with less effort and by the usually available clinically experienced staff. It should be mentioned in this context that imaging and spectroscopic studies on lipid depositions cannot replace one another, since they are focusing on different organs and metabolic pathways. A combination of a variety of techniques including the new standardized imaging approach will further elucidate the interactions of different lipid compartments in the body and their regulation under normal and pathological conditions.