Separation of Abdominal Subcutaneous Adipose Tissue (SAT) and Visceral Adipose Tissue (VAT) based on Wheel-Template in MRI
Steve Cheuk Ngai Hui1, Teng Zhang1, Defeng Wang1, and Winnie Chiu Wing Chu1

1Dept. of Imaging and Interventional Radiology, The Chinese University of Hong Kong, Hong Kong, Hong Kong


This abstract introduces the use of wheel-template to perform segmentation on subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT) at the abdominal region. The proposed method detects narrow regions between SAT and VAT, and uses line cut to separate two types of tissues based on MRI data. It performs well on obese individuals and the obtained results are correlated to those obtained from a semi-automatic method. A quantitative measurement of SAT and VAT is important as they are developed from different pathways and suggested to be related to different chronic diseases.


Several methods have been proposed to segment subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT) based on active contour algorithm, geometrical model, registration based, graph cut, thresholding and intensity level (1-6). Nevertheless, morphological or deformational based segmentation is challenging to obtain high accuracy due to high variations of abdominal shape especially at the pelvis regions. The proposed study suggests a different approach by cutting the boundary between SAT and VAT using circular templates and line-cut that satisfied the given constraints to detect the narrow boundaries.


Subjects: Fourteen obese adolescents (Female: 9, Male: 5, age: 17.3±1.4, BMI: 33.2±3.6) under the health monitoring program were recruited from the out-patient clinic at the university teaching hospital. Inclusion criteria included age between 13-19 years, BMI > 25, and had no long term diseases affecting the weight. Image acquisition: The abdominal regions of all subjects were scanned using a 3.0 T whole-body scanner (Achieva X-series, Philips Medical System, Best, The Netherlands) with a 16-channel SENSE-XL-Torso array coil. The abdominal regions covered from the dome of diaphragm to the pubic symphysis and divided into three sections during the scan. Chemical-shift water-fat images were acquired using a 3D multi echo DIXON sequence (TR = 13 ms, TE1/TE2 = 1.2/2.2 ms, flip angle = 3°, FOV = 450 (right-left) x 340 (AP) mm, slice thickness = 10 mm, number of slices = 30) to yield co-registered water, fat, in-phase, and opposed-phase image series. Image preprocessing: An in-house segmentation algorithm was developed on Matlab (R2011a, Mathworks, Natick, USA) to separate SAT and WAT at the abdominal regions. Smoothing and de-noising were performed on the Fat-only series images followed by thresholding using Otsu's method to create a binary mask (7,8). The arms of the subjects were removed by separating from the largest connected regions. Line-Cut: To separate the connected regions between SAT and VAT (figure 1), a circular mask was created using Midpoint circle algorithm followed by adding straight lines on the mask through the center using Bresenham's line algorithm. The numbers of lines depended on the diameter of the circle. The adipose tissues (white pixels) were scanned with the template, and if both end points of any line on the circular mask touched non-adipose tissues (dark pixels), the coordination set of the line points were labeled and a black line would be drawn after the scan (figure 2). SAT was segmented by getting the largest connected component and VAT was obtained by subtracting the SAT from the binary mask as shown in figure 3. Validation: Results from the proposed method were compared with the results obtained using a semi-automatic segmentation on ITK-SNAP 2.40 (9). Pearson Correlation Coefficient was measured to test the relationship between two sets of results.


Results are presented in VAT, SAT and total adipose tissue (TAT) as shown in table 1.Pearson correlation coefficient indicates that results from the proposed method are significantly correlated with those from the semi-automatic method.


The proposed method is able to work on abdominal regions with large shape variation. It overcomes the challenges in other methods based on active contour and deformable model. It detects narrow connecting parts using a wheel-template (circular template) with lines through the center. The proposed method is also less sensitive to noise, motion artifacts and signal inhomogeneity as it only detects narrow boundaries between SAT and VAT. The volume of SAT would be slightly underestimated as a small number of pixels at the outer and inner boundary are removed due to the process of the line cut algorithm when the template is scanning at the edge. These pixels are presented in VAT results as shown in the small white pixels along the boundary in figure 3c. VAT volume would be overestimated on the other hand due to the edge pixels from SAT and the femur bones and joints at the pelvis area. TAT results are very close in both methods due to using the same raw images and the same thresholding method to measure the adipose tissues. Testing on normal subjects will be conducted in the near future and one potential problem would be on very thin individuals with a small thickness of SAT which is thinner than the diameter of the predefined circular template. In conclusion, as SAT and VAT are different in many aspects, an accurate and quantitative measurement allows physicians to evaluate the efficacy of treatments against obesity and the relationship between diseases to different kinds of adipose tissues.


The work described in this paper was fully supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No. SEG CUHK_02)


1. Kullberg J, Karlsson AK, Stokland E, Svensson PA, Dahlgren J. Adipose tissue distribution in children: automated quantification using water and fat MRI. J Magn Reson Imaging 2010;32(1):204-210.

2. Kullberg J, Ahlstrom H, Johansson L, Frimmel H. Automated and reproducible segmentation of visceral and subcutaneous adipose tissue from abdominal MRI. Int J Obes (Lond) 2007;31(12):1806-1817.

3. Positano V, Gastaldelli A, Sironi AM, Santarelli MF, Lombardi M, Landini L. An accurate and robust method for unsupervised assessment of abdominal fat by MRI. J Magn Reson Imaging 2004;20(4):684-689.

4. Wald D, Teucher B, Dinkel J, et al. Automatic quantification of subcutaneous and visceral adipose tissue from whole-body magnetic resonance images suitable for large cohort studies. J Magn Reson Imaging 2012;36(6):1421-1434.

5. Joshi AA, Hu HH, Leahy RM, Goran MI, Nayak KS. Automatic intra-subject registration-based segmentation of abdominal fat from water-fat MRI. J Magn Reson Imaging 2013;37(2):423-430.

6. Sadananthan SA, Prakash B, Leow MK, et al. Automated segmentation of visceral and subcutaneous (deep and superficial) adipose tissues in normal and overweight men. J Magn Reson Imaging 2015;41(4):924-934.

7. Coupe P, Manjon JV, Gedamu E, Arnold D, Robles M, Collins DL. Robust Rician noise estimation for MR images. Med Image Anal 2010;14(4):483-493.

8. Otsu N. A Threshold Selection Method from Gray-Level Histograms. Systems, Man and Cybernetics, IEEE Transactions on 1979;9(1):62-66.

9. Yushkevich PA, Piven J, Hazlett HC, et al. User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability. NeuroImage 2006;31(3):1116-1128.


Figure 1: Regions that connected SAT and VAT highlighted in red circle.

Figure 2: a conceptual figure that shows the wheel-template which both end-points of the red lines touched the non-adipose tissue (dark pixel). The red line indicated would be replaced by black pixel to perform a cut on the graph. The radius of the template is adjustable in the algorithm.

Figure 3 (a) original binary mask, (b) results of SAT, (c) results of VAT.

Table 1: Volume of adipose tissue from the proposed method compared with the semi-automatic method.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)