Using a phantom to compare MR techniques for determining the ratio of intraabdominal to subcutaneous adipose tissue

Patients who have a greater distribution of intraabdominal adipose tissue as compared with subcutaneous adipose tissue and an increased ratio of intraabdominal adipose tissue to subcutaneous adipose tissue are at greater risk for developing cardiovascular disease and type 2 diabetes mellitus. In previous MR investigations, researchers have used conventional T1-weighted spin-echo images to determine the ratio of intraabdominal adipose tissue to subcutaneous adipose tissue. However, no investigation, to our knowledge, has been performed to determine the accuracy of using different MR sequences to estimate adipose distribution. The purpose of our investigation was to compare MR imaging and segmentation techniques in calculating the ratio of intraabdominal to subcutaneous adipose tissue using an adiposity phantom.

A phantom was created to simulate the distribution of subcutaneous and intraabdominal fat (with known volumes). Axial MR images were obtained twice through the phantom using a 5-mm slice thickness and zero gap for the following T1-weighted sequences: spin-echo, fast Dixon, and three-dimensional (3D) spoiled gradient-echo. An in-house computer software program was then used to segment the volumes of fat and calculate the volume of intraabdominal adipose tissue and subcutaneous adipose tissue and the ratio of intraabdominal to subcutaneous adipose tissue. Each imaging data set was segmented three times, so six sets of data were yielded for each imaging technique. The percentage predicted of the true volume was calculated for each MR imaging technique for each fat variable. The mean percentages for each variable were then compared using one-factor analysis of variance to determine whether differences exist among the three MR techniques.

The three MR imaging techniques had statistically significant different means for the predicted true volume of two variables: volume of subcutaneous adipose tissue (p < 0.001) and volume of intraabdominal adipose tissue (p = 0.0426). Estimates based on fast Dixon images were closest to the true volumes for all the variables. All MR imaging techniques performed similarly in estimating the ratio of intraabdominal adipose tissue to subcutaneous adipose tissue (p = 0.9117). The acquisition time for the 3D spoiled gradient-echo images was 10-22 times faster than for the other sequences.

Conventional T1-weighted spin-echo MR imaging, the current sequence used in practice for measuring visceral adiposity, may not be the optimal MR sequence for this purpose. We found that the T1-weighted fast Dixon sequence was the most accurate at estimating all fat volumes. The T1-weighted 3D spoiled gradient-echo sequence generated similar ratios of intraabdominal to subcutaneous adipose tissue in a fraction of the acquisition time.