Abdominal Adipose Tissue Distribution on MRI and Diabetes

Abdominal Adipose Tissue Distribution on MRI and Diabetes Michalis Mantatzis, MD, PhD, Thanos Milousis, MD, Simoni Katergari, MD, Andreas Delistamatis, PhD, Dimitrios N. Papachristou, MD, PhD, Panos Prassopoulos, MD, PhD Rationale and Objectives: To introduce a simple magnetic resonance imaging (MRI) protocol for quantitative assessment of intraperitoneal, retroperitoneal, and subcutaneous adipose tissue (AT) and to compare AT distribution between diabetic and nondiabetic individuals. Materials and Methods: Thirty-eight consecutive male diabetic patients (group A) and 38 males (who matched for body mass index [BMI]) without metabolic syndrome (group B) underwent abdominal MRI with a three-dimensional spoiled gradient echo T1-weighted sequence. The amounts of intraperitoneal, retroperitoneal, and subcutaneous AT were calculated on a workstation, after manual anatomic segmentation and were correlated with 10 anthropometric measurements. Pearson product–moment correlation coefficients were used for correlation of AT volumes with anthropometric measurements, Wilcoxon test to compare AT measurements between automatic and manual technique used, and unpaired t test to compare volumes of AT compartments between group A and B. Results: Diabetic patients exhibited larger amount of intraperitoneal and retroperitoneal AT than normal individuals at all levels (t = 2.02,P < .05). Among anthropometric measurements, the waist circumference, BMI, and body fat percentage exhibited the best correlations with intraperitoneal and retroperitoneal AT (group A (r) = 0.88/0.78/0.0.69 and group B (r) = 0.91/0.87/0.81). The L2–L5 set of images was found to be the most representative of the amount of AT volumes. Conclusions: Amount and distribution of AT can be accurately and easily assessed on MRI. Quantification of intraabdominal AT may promote the role of imaging in the study of metabolic syndrome. Key Words: Adipose tissue; MRI; segmentation, body fat measurements; metobolic syndrome; diabetes. ªAUR, 2014

O

besity has emerged as one of the most serious public health concerns in the 21st century in both developed and developing countries, because it is strongly related with the modern dietary habits and lifestyle. Increased body mass index (BMI) and especially increased visceral fat have been recognized as the major contributors to metabolic disturbances and the development of diseases including type 2 diabetes, cardiovascular disease, hypertension, and stroke. Quantification of obesity and adipose tissue (AT) distribution within the body is fundamental for the evaluation of patients with diabetes and metabolic syndrome. It is usually performed by simple anthropometric measurements, including BMI for rough estimation of total body fat, waist circumference (WC) for visceral AT (VAT), hip circumference for subcutaneous AT (SAT), and waist-to-hip ratio (WHR) (1). Although these measurements can be easily acquired, they are not precise, they may suffer from systematic errors (2),

and they cannot accurately assess the most crucial AT compartment, that is, the VAT, proven to be closely related with the development of metabolic syndrome (3). Alternatively, computed tomography (CT) and magnetic resonance imaging (MRI) have been used for direct (in vivo) measurements of total body AT and of the major AT compartments, that is, visceral and subcutaneous AT (4–7). Various strategies have been proposed for AT assessment using imaging, including measurements of head-to-toes body fat, those of whole abdominal fat, or those of fat on a single-slice CTor MRI at different predetermined levels in the abdomen (4,8–11). However, there is no consensus on the most appropriate measurement methodology on imaging examinations. The purpose of this study was to introduce a simple MRI protocol for the assessment of AT distribution in the body, to evaluate differences in AT distribution between diabetic patients and nondiabetic subjects on MRI, and to correlate common anthropometric measurements with MRI assessment of visceral and subcutaneous AT.

Acad Radiol 2014; 21:667–674 From the Department of Radiology, University Hospital of Alexandroupolis, Democritus University of Thrace, Dragana, Alexandroupolis 68100, Greece (M.M., A.D., P.P.); Department of Internal Medicine and Endocrinology (T.M., S.K.); and Department of Endocrinology, University Hospital of Alexandroupolis, Democritus University of Thrace, Alexandroupolis, Greece (D.N.P.). Received November 15, 2013; accepted January 12, 2014. Address correspondence to: M.M. e-mail: [email protected] ªAUR, 2014 http://dx.doi.org/10.1016/j.acra.2014.01.009

MATERIALS AND METHODS Patients

For the purposes of this study, 76 patients were gathered in two groups. Group A enrolled 38 consecutive men with type 2 diabetes mellitus, (mean age = 60.1, range = 44–71) 667

MANTATZIS ET AL

Academic Radiology, Vol 21, No 5, May 2014

Figure 1. (a) High–signal intensity adipose tissue (AT) is depicted in the subcutaneous, intraperitoneal, and retroperitoneal compartments. Fat signal intensity pixels are labeled after thresholding and measured in compartments outlined by the operator. (b) Abdominal subcutaneous AT compartment (in red) is clearly separated from visceral fat and intramuscular fat (in blue). Visceral, that is, intraperitoneal and retroperitoneal (c) and retroperitoneal (d), AT compartments are separated with a line running posterior to ascending colon, small intestine, and descending colon, anterior to the pancreas and great vessels. (Color version of figure is available online.)

with newly diagnosed disease, according to American Diabetes Association (ADA) criteria (12). The control group (group B) was prospectively recruited from BMI-matched male patients who underwent abdominal MRI with various indications, and no significant abnormality was detected. Group B consisted of 38 men (mean age = 54.44, range = 38–72). Patients had no clinical or laboratory evidence of metabolic disease, and MRI findings were unremarkable. Informed consent was obtained from all patients. Anthropometric Measurements

Anthropometric measurements were acquired by the same clinician (T.M.) and included weight, height, BMI, waist and hip circumferences, WHR, femur length, waist-tofemur ratio, waist-to-height ratio, arm span, and skin fold thickness. Body weight was calculated to the nearest 0.1 kg with patients wearing light clothing and no shoes, and height was calculated to the nearest 1 cm. Waist and hip circumferences of subjects in standing position were measured using a soft tape. Maximal WC was measured between the lower rib margin and the iliac crest. Hip circumference was defined as the maximal body perimeter over the buttocks. Arm span was measured with the individual in standing position having both hands in abduction, and femur length was measured in the anterior surface from anterior superior iliac spine to the center of patella. Thickness of triceps skin fold was measured using a suitable instrument. Waist-to-hip, waist-to-height, and waist-to-femur ratios were calculated from the derived relevant data. BMI was calculated using the formula, BMI = height/weight (2), expressed in kg/m2. Finally, body fat accumulation was estimated by bioimpedance analysis using OMRON BF300 Body Fat Monitor (Matsukasa OMRON Co. LTD, Japan). 668

Magnetic Resonance Imaging

All subjects underwent an MRI examination on a Signa Horizon LX 1.0 T magnetic resonance unit (GE Medical Systems, Milwaukee, WI), in a supine position using a body coil. A three-dimensional T1 spoiled gradient echo (SPGR) sequence was obtained through the abdomen (repetition time [TR] = 9.4–9.5 milliseconds, echo time [TE] = 1.9–2.0 milliseconds, flip angle = 30 , field of view [FOV] = 380 380– 480 480 mm, and slice thickness = 8 mm) and axial T1-weighted Fast Spin Echo [FSE] acquisition at the lower pelvis–hip levels (TR = 340, TE = 12, slice width = 6 mm, FOV = 380 380–480 480 mm) were acquired and transferred to a PC workstation for AT measurements. Both sequences required for the study were performed in 4 minutes. Image Analysis

Each image was processed by dedicated software developed on purpose for semiautomatic measurement of AT. The software processed Digital Imaging and Communications in Medicine (DICOM) images after conversion to 8-bit gray-scale images. Each pixel was labeled as fat or nonfat by applying a threshold to the brightness of each pixel. ‘‘Bright’’ pixels that corresponded to AT were displayed in red (Fig 1). The anatomic regions where the ATwas to be estimated, namely subcutaneous, visceral, and retroperitoneal, were outlined jointly by two examiners (P.P., M.M.), with more than 15 and 5 years experience in abdominal imaging, respectively, by manual tracing of the appropriate compartments. To measure the subcutaneous AT, the external margin of the abdominal muscles was traced (Fig 1b), AT pixels outside tracing were automatically labeled—calculated and the results were expressed as surface (mm2) and volume (mm3) using the


ABDOMINAL AT DISTRIBUTION ON MRI AND DIABETES

DICOM information of the sequence (pixel size and slice thickness). Visceral (intraperitoneal and retroperitoneal) AT pixels were labeled—calculated after tracing of the inner margin of abdominal/back muscles on all abdominal images (Fig 1c). Retroperitoneal AT was defined as the AT located behind a line running posterior to ascending colon and small intestine, anterior to the pancreas and great vessels, and posterior to the descending colon (Fig 1d). At the hips, the outer border of abdominal and gluteal muscles was traced, and the pixels outside tracing were calculated as subcutaneous AT (Fig 2). Tracing and calculation were made at the slice where the maximum amount of SAT was found at the hip level. The AT volume for each anatomic compartment was recorded in each slice from the abdomen and the hip level. The sum of these volumes was referred as ‘‘total AT’’ for each selected slice. Subsequently, the average visceral, subcutaneous, and total fat were calculated from the whole set of images. The average measurements, the measurements at the hip level, and the L2–L3, L3–L4, and L4–L5 intervertebral disk levels—selected using the scanogram—were transferred to the logistic spreadsheet for the statistical analysis. To evaluate the accuracy of our software in measuring fat tissue volumes, we performed measurements of fat using summation of areas technique in 15 patients on a workstation (Centricity; GE Medical Systems, Milwaukee, WI), and we subsequently estimated the agreement with the proposed semiautomated method. Statistics

Data were assembled in an Excel 2010 spreadsheet (Microsoft Corp, Seattle, WA), were statistically analyzed using the Statistical Package for the Social Sciences v 17.0 (SPSS Inc., Chicago, IL), and were presented as group mean values standard deviation (SD). AT volumes calculated using MRI were correlated with anthropometric measurements using Pearson product–moment correlation coefficients. To evaluate the accuracy of the specific software in measuring AT compartments, we used the Wilcoxon nonparametric test between automatic and manual measurements. After check for normal distribution, unpaired t test was applied to compare the mean volumes of AT compartments between groups A and B. The power of the used statistic methods was assessed with study size 2.0 (CreoStat, Frolunda, Sweden.) The study was conducted in accordance to the Declaration of Helsinki, and all participants received both oral and written information and signed an approved informed consent form before entering the study. The MRI protocol is used in regular magnetic resonance (MR) examinations in everyday clinical practice.

RESULTS Table 1 presents the mean values SD of anthropometric measurements and of AT volumes measured using MRI in different anatomic compartments and selected levels, separately for group A and group B.

Figure 2. Axial T1-weighted Fast Spin Echo (FSE) at the hip level. Pixels representing adipose tissue (AT) are bright (a), and they are labeled with red color after thresholding (b). After manual anatomic segmentation (c), subcutaneous fat pixels are measured, whereas intraabdominal AT as well as intramuscular and bone marrow AT are excluded. (Color version of figure is available online.)

Diabetic patients exhibited a larger amount of VAT than the control group at all levels, (average VAT [t test = 2.184, P = .049] and VAT at the L3–L4 level [t test = 2.002, P = .043]). Similarly, diabetic patients had increased amount of retroperitoneal AT than control group at L2–L3 level (t test = 2.81, P = .005) and at L3–L4 level (t test = 2.26, P = .027). Subcutaneous AT did not differ, at any level, between the two groups on MRI. Correlations between anthropometric measurements and AT measurements on MRI are presented in Table 2. WC 669

MANTATZIS ET AL


TABLE 1. Anthropometric Measurements—Descriptive Statistics

Weight (kg) Height (cm) Body mass index (kg/m2) Waist circumference (cm) Hips circumference (cm) WHR Femur (cm) Waist-to-height ratio Waist-to-thigh ratio Arm span Skin fold (cm) Abdominal VAT (mm2) Intraperitoneal (mm2) Retroperitoneal (mm2) Abdominal SAT (mm2) Abdominal SAT + VAT (mm2) Hip SAT (mm2)

Diabetics (n = 38)

Normal (n = 38)

88.68 13.41 167 5.6 31.51 4.45 107.89 10.16 105.64 7.42 1.02 0.05 62.33 7.89 0.64 0.05 1.74 0.17 173.9 7.23 15.16 7.04 22229.03 1379.96 13026.36 1014.37 8849.47 635.39 21148.07 1462.16 43377.1 2275.18 20666.37 1288.67

88.1 15.7 171 6.7 30.02 4.7 104.1 12.3 105 8.98 0.98 0.05 64.04 10.56 0.6 0.06 1.67 0.47 177.94 7.34 15.13 12.0 18915.24 1246.75 11598.82 763.97 7316.42 532.57 20893.2 1673.74 39808.45 2347.78 19474.35 1536.47

SAT, subcutaneous adipose tissue; VAT, visceral adipose tissue, WHR, waist-to-hip ratio. Results are presented as mean standard deviation.

exhibited the best correlation with the sum of abdominal AT (visceral + subcutaneous, group A [r] = 0.88, group B [r] = 0.91), with AT in L2–L3 level in diabetic patients (r = 0.88) and with AT in L4–L5 level in the control group (r = 0.93). Similarly, Hip Circumference (HC) exhibited the best correlation with AT at the hips measured on MRI (group A [r] = 0.88, group B [r] = 0.87). However, WHR showed poor-to-moderate correlation with AT measurements on MRI (0.33 < r < 0.736, .001 < P < .05) and poor correlation with the VAT-to-SAT ratio on MRI (group A [r] = 0.36, group B [r] = 0.44, P < .001). BMI demonstrated the best correlation with AT at L2–L3 level (r = 0.83) in diabetic patients and AT at L4–L5 level (r = 0.9) in the control group. BMI and sum of abdominal AT exhibited significant correlation in both groups (group A [r] = 0.78, group B [r] = 0.87). Body fat percentage most significantly correlated with the sum of abdominal AT (group A [r] = 0.69, group B [r] = 0.81), with AT in L2–L3 level in diabetic patients (r = 0.72) and with L4–L5 in normal individuals (r = 0.82). The L2–L5 VAT volume always exhibited correlation close to the highest correlation levels observed and better than the rest single levels. Correlations of the rest anthropometric measurements with measurements of AT on MR images were either statistically insignificant or exhibited weak correlations. The dedicated software for AT measurements provided similar results with manual measurements performed on the workstation; there were no statistical differences between the two sets of measurements (Wilcoxon z = 1.44, P = .14), and the differences of mean values were