Kidney segmentation in 4-dimensional dynamic contrast- enhanced MR images : A physiological approach

Abstract

Four-dimensional dynamic contrast enhanced-magnetic resonance imaging (4-D DCE-MRI) is an emerging clinical tool in paediatric urology. It involves administration of a contrast agent whose journey through the kidneys is monitored over time.This thesis proposes two types of techniques to segment kidney tissues in 4-D DCE-MRI by exploiting the kinetics of contrast agent uptake in abdominal tissues. A clustering based technique identifies groups of voxels that share similar time courses and associates each group to a different tissue type. A factor analysis based technique (FAMIS) with novel constraints yields factors that describe physiologically both the spatial distribution and temporal behaviour of the different tissue types.The techniques are validated over a wide range of 20 patient MR data sets and results show good agreement with the manual segmentation performed by the expert. Especially, the FAMIS technique has the potential to produce physiologically interpretable factors.