Harald F. Busse PhD, Presenter: Nothing to Disclose

Alexander Schaudinn MD, Abstract Co-Author: Nothing to Disclose

Nicolas Linder, Abstract Co-Author: Nothing to Disclose

Gregor Thormer, Abstract Co-Author: Employee, Siemens AG

Thomas Kurt Kahn MD, Abstract Co-Author: Nothing to Disclose

Nikita Garnov, Abstract Co-Author: Nothing to Disclose

The presented software enables visualization and quantification of various fat depots and is considered a valuable tool to assess disease conditions and monitor related interventions.

With obesity-related diseases, such as type 2 diabetes, on the rise, quantification of visceral and subcutaneous adipose tissue (VAT, SAT) volumes is becoming increasingly important as a diagnostic means for risk assessment. MRI-based analysis is common for that purpose but is either time-consuming with manual or error prone with automatic data processing. We report on our advanced experience, highlighting benefits and limitations of a customized semiautomatic fat quantification tool that has been used over the last three years for VAT and SAT analysis in obese patients.

The Matlab tool works with Dixon MR images, at our site, with 2-point acquisitions in supine position (1.5 T Achieva XR, Philips, 50 slices, 10 mm thick, 0.5 mm gap, in 160 s plus breathing intervals). An active contour model is used to define inner and outer VAT and SAT boundaries. VAT volumes are quantified by histogram analysis of the MR signal intensities. Starting at an automatic threshold, the user has immediate visual feedback of the segmented VAT image as the threshold is adjusted until results are acceptable. Also, SAT and VAT outlines can easily be corrected manually. Work can be saved and retrieved at any time for later processing. SAT and VAT total volumes and per slice are reported in common spreadsheet format.

This tool has been used on over 500 datasets, originally covering 20 slices in the lumbar region and, for more than a year, ≈ 40 slices across the whole abdomen. About 1 in 6 slices require minor and another 1/6 major corrections. Mean segmentation time for total VAT is ≈ 24 min. Difficulties in automatic segmentation arise, e. g., from liver or intestinal fat that is mistaken for VAT, fatty abdominal muscles with tissues mixed, a limited FOV or artifacts occurring for BMIs > 40, and in regions like the minor pelvis or diaphragm dome where specific fat signals are missing. On the other hand, 4 in 6 slices can be left as is, and corrections for patients with intact abdominal muscle layers are minimal.

Busse, H, Schaudinn, A, Linder, N, Thormer, G, Kahn, T, Garnov, N, Advanced Experience with a Semi-automatic, Customized Software Tool for Clinical MRI Quantification of Visceral and Subcutaneous Adipose Tissue.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14018565.html