Topography mapping of whole body adipose tissue using A fully automated and standardized procedure

• Published in: Journal of magnetic resonance imaging 2010-02, Vol.31 (2), p.430-439
• Main Authors: Würslin, Christian, Machann, Jürgen, Rempp, Hansjörg, Claussen, Claus, Yang, Bin, Schick, Fritz
• Format: Article
• Language: English
• Subjects: Adipose Tissue - anatomy & histology; adipose tissue profiling; Algorithms; Artificial Intelligence; automatic image segmentation; Body Fat Distribution - methods; Body Fat Distribution - standards; Germany; Humans; Image Enhancement - methods; Image Enhancement - standards; Image Interpretation, Computer-Assisted - methods; Image Interpretation, Computer-Assisted - standards; Imaging, Three-Dimensional - methods; Imaging, Three-Dimensional - standards; Magnetic Resonance Imaging - methods; Magnetic Resonance Imaging - standards; metabolism; Pattern Recognition, Automated - methods; Reproducibility of Results; Sensitivity and Specificity; visceral adipose tissue; Whole Body Imaging - methods; Whole Body Imaging - standards; whole body MRI
• ISSN: 1053-1807; 1522-2586
• DOI: 10.1002/jmri.22036

Purpose: To obtain quantitative measures of human body fat compartments from whole body MR datasets for the risk estimation in subjects prone to metabolic diseases without the need of any user interaction or expert knowledge. Materials and Methods: Sets of axial T1‐weighted spin‐echo images of the whole body were acquired. The images were segmented using a modified fuzzy c‐means algorithm. A separation of the body into anatomic regions along the body axis was performed to define regions with visceral adipose tissue present, and to standardize the results. In abdominal image slices, the adipose tissue compartments were divided into subcutaneous and visceral compartments using an extended snake algorithm. The slice‐wise areas of different tissues were plotted along the slice position to obtain topographic fat tissue distributions. Results: Results from automatic segmentation were compared with manual segmentation. Relatively low mean deviations were obtained for the class of total tissue (4.48%) and visceral adipose tissue (3.26%). The deviation of total adipose tissue was slightly higher (8.71%). Conclusion: The proposed algorithm enables the reliable and completely automatic creation of adipose tissue distribution profiles of the whole body from multislice MR datasets, reducing whole examination and analysis time to less than half an hour. J. Magn. Reson. Imaging 2010; 31: 430–439. © 2010 Wiley‐Liss, Inc.