Accurate Adipose Tissue Segmentation from Single Gradient Echo Phase Images by Adaptive Local Thresholding

Christian Wrslin¹, Frank Eibofner¹, Fabian Springer¹, Fritz Schick¹

In many applications, a quantification of fat is desired. Most (semi-)automatic procedures use T1-weighted spin echo images to accomplish this. These approaches are time-consuming and a precise quantification is usually complicated by a high amount of partial volume effects. We propose a quantification procedure based on one single phase image, acquired with a gradient echo technique and opposed-phase condition. This maximizes the contrast in between fat- and water- dominated tissues and is less time-consuming. The phase divergence, arising from inhomogeneities, is compensated for using an automated algorithm, enabling a precise fat quantification by thresholding. Phantom measurements show a high precision.