Image analysis methods for dynamic hepatocyte-specific contrast enhanced MRI

Abstract

Patients with colorectal cancer are frequently presented with liver metastases for which (partial) resection is often the best therapy. However, the future remnant liver, the remaining part of the liver after resection, should allow adequate liver function to avoid liver failure. This thesis presents novel methods for the accurate voxel-wise estimation of the future remnant liver’s function based on pharmacokinetic modeling of dynamic contractenhanced (DCE)MRI. The methods comprise a variety of novel techniques for DCE-MRI of the liver: 1) 4D registration of the DCE series; 2) delineation of the liver, the liver vasculature and the liver’s anatomical segments; 3) pharmacokinetic (PK) modeling of the perfusion based on the intra-cellular contrast agent Gd-EOB-DTPA (Primovist); 4) assessment of the relation between DCE-MRI and hepatobiliary scintigraphy (HBS). Spatial alignment of the voxels in the 4D DCE-MRI is an important requirement for PK modeling. We exploit the proximity of deformation fields to sequentially register images in an ordered fashion. The global liver displacement helps in predicting the deformation ‘tendency’ along the time axis. The deformation tendency allows us to obtain a better starting point for the registration. Such a method aims to start the registration optimization close to the optimum and avoid getting trapped in a local minimum. We apply a liver-specific contrast agent, due to which the liver shows