Abstract
Functional studies are routinely performed to provide quantitative evaluations of renal function. Of these, Nuclear Radioisotope Renography is the current technology of choice in clinical studies. However, there has been considerable interest in using Magnetic Resonance (MR) in these studies, motivated by avoidance of the use of ionising radiation, which is particularly attractive for paediatric subjects and by the availability of high quality anatomical and functional information within a single investigation, thus allowing simultaneous observation of pathology and renal function. The overarching factor preventing widespread use of MR Renography is the lack of image processing techniques capable of robust extraction of quantitative information from the available data. The work presented in this thesis is aimed at addressing this issue. A methodology for estimating renal function is proposed. Movement correction of Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE-MRI) data was performed using a 3D movement correction method based on the phase difference of time-adjacent data volumes. This was followed by a novel approach demonstrating the first Partial Volume (PV) Effect correction on MR Renography data. For a typical renal Region of Interest, the observed intensity for each voxel was de-composed into its constituent parts, corresponding to the contributions from each tissue, using knowledge of the Point Spread Function (PSF) and high-resolution registered templates for each anatomical tissue type. Thus, non-renal contributions from liver, spleen and other surrounding tissues could be eliminated from time-intensity curves. This produced a change the renal curve that resulted in enhanced Glomerular Filtration Rate (GFR) estimates, as per a Rutland-Patlak analysis of the time intensity data: a cohort of 10 healthy volunteers produced a mean enhancement of 36% in relative GFR with a mean improvement of 5% in R2 fitting the Rutland-Patlak model when PV correction was applied compared to no PV correction. PV correction in small structures (with respect to the PSF) corresponding to arterial vasculature were also investigated using a statistical Partial Volume classifier and a novel mixing prior that models more closely asymmetries in intensity distributions related to small object size than others proposed in the literature. Classifier performance was found to be approximately inversely proportional to Contrast to Noise. An intensity recovery step was also proposed for Arterial Input Functions derived from such small objects. The resolution of the PSF of the DCE-MRI sequence was found to similar to typical aortic diameters and therefore would present intensity reductions due to the PV Effect of ~ 60% that need to be recovered. This work has shown that in addition to movement correction, contrast agent quantification and accurate modelling, absolute GFR measurement via DCE-MRI is significantly affected by the PV Effect, and as such must be robustly accounted for.