Automated segmentation of adipose tissue in magnetic resonance images of the whole body

The amount and the bodily distribution of different adipose tissue (AT) compartments are important indicators for the risk of obesity-related diseases and play an important role in the investigation of their pathogenesis. Magnetic resonance imaging can be used to acquire images of the whole body, showing these compartments and their distribution. In this article, an automated segmentation algorithm is presented, being able to create tissue profiles of the whole body for tissue classes subcutaneous AT, visceral AT and total tissue. The images are segmented using a fuzzy c-means algorithm, which considers partial volume effects. A separation of the body into anatomic regions along the body axis is done to define regions with visceral AT present. In abdominal image slices, the AT compartments are divided into subcutaneous and visceral compartments using an active contour algorithm. The slice-wise areas of different tissues are plotted against the slice position to obtain their topography. The automatically obtained tissue profiles were compared to profiles created manually by an expert and show high correlation coefficients, indicating similar topography. Absolute error values were calculated for evaluation of the algorithm's absolute accuracy. These show low overall mean values for the classes of total tissue (4.48%) and visceral AT (3.26%). The deviation of total AT (sum of visceral and subcutaneous AT) was higher though (8.71%). Whole examination and analysis time is reduced to less than half an hour.