ABSTRACT
PURPOSE: To assess the feasibility of multiple-bolus dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) in the pancreas; to optimize the analysis; and to investigate application of the method to a glucose challenge in type 2 diabetes. MATERIALS AND METHODS: A 4-bolus DCE-MRI protocol was performed on five patients with type 2 diabetes and 11 healthy volunteers during free-breathing. Motion during the dynamic time series was corrected for using a model-driven nonlinear registration. A glucose challenge was administered intravenously between the first and second DCE-MRI acquisition in all patients and in seven of the healthy controls. RESULTS: Image registration improved the reproducibility of the DCE-MRI model parameters across the repeated bolus-acquisitions in the healthy controls with no glucose challenge (eg, coefficient of variation for K(trans) improved from 38% to 28%). Native tissue T(1) was significantly lower in patients (374 +/- 68 msec) compared with volunteers (519 +/- 41 msec) but there was no significant difference in any of the baseline DCE-MRI parameters. No effect of glucose challenge was observed in either the patients or healthy volunteers. CONCLUSION: Multiple bolus DCE-MRI is feasible in the pancreas and is improved by nonlinear image registration but is not sensitive to the effects of an intravenous glucose challenge.
Subject(s)
Contrast Media , Diabetes Mellitus, Type 2/diagnosis , Gadolinium DTPA , Magnetic Resonance Imaging/methods , Pancreas/anatomy & histology , Adult , Aged , Blood Glucose/analysis , Case-Control Studies , Female , Glucose Tolerance Test , Humans , Male , Middle Aged , Pulse Therapy, Drug , Reproducibility of Results , Sensitivity and SpecificityABSTRACT
This paper presents a series of 3D statistical models of the cortical sulci. They are built from points located automatically over the sulcal fissures, and corresponded automatically using variants on the iterative closest point algorithm. The models are progressively improved by adding in more and more structural and configural information, and the final results are consistent with findings from other anatomical studies. The models can be used to locate and label anatomical features automatically in 3D MR images of the head, for analysis, visualisation, classification, and normalisation.
