Automated quantification of muscle and fat in the thigh from water-, fat- and non-suppressed MR images

A key parameter in metabolic and pathologic studies is the estimation of body tissue distribution. This is a laborious and operator-dependent process. In this work we introduce an unsupervised muscle and fat quantification algorithm based on water only, fat only and water-and-fat MRI images of the mid-thigh area. We first use parametric deformable models to segment the subcutaneous fat and then apply centroid clustering in the feature domain defined by the voxel intensities in water only and fat only images to detect the inter-muscular fat, muscle and bone. This tissue decomposition permits the computation of volumetric and area measures of fat and muscle. We tested the proposed method on 9 participants and validated these measures against values obtained from a semi-manual clinician-driven analysis of single-slice mid-thigh CT images of the same participants. Our approach was found to be statistically consistent with the semi-manual reference method, and was able to address inter-participant anatomic variability and intensity inhomogeneity effects.