Search

Sinusoidal capillarization induces microcirculatory changes in liver cirrhosis and fibrosis. The purpose of this study was to assess whether contrast-enhanced MRI can be used to demonstrate the effects of sinusoidal capillarization in liver fibrosis. Dynamic MRI after injection of a low-molecular-weight contrast agent of 0.56 kDa (Gd-DOTA), and two high-molecular-weight contrast agents of 6.47 kDa and 52 kDa (P792 and P717) was performed in rabbits with liver fibrosis induced by cholesterol and diethylstilbestrol. The hepatic distribution volume accessible to the high-molecular-weight agents decreased in the rabbits with liver fibrosis (P792: 7.8% +/- 1.7% vs. 10.1% +/- 1.8% in normal rabbits, P =.038; P717: 6.2% +/- 2.1% vs. 9.7% +/- 1.6% in normal rabbits, P =.007), whereas the hepatic mean transit time (MTT) of the low-molecular-weight agent was increased (15.9 +/- 8.0 s vs. 8.8 +/- 2.6 s in normal rabbits, P =.015). In rabbits with liver fibrosis, the clearance of indocyanine green (ICG) was correlated with the volume accessible to the high-molecular-weight agents (P792: r = 0.810, P =.015; P717: r = 0.857, P =.007). The collagen content of the liver was inversely correlated with the distribution volume of P717 (r = -.833, P =.010) and with the ICG clearance (r = -.810, P =.015). It was concluded that the microcirculatory changes induced by sinusoidal capillarization in liver fibrosis can be demonstrated noninvasively with MRI.

The aim of this study was to assess the importance of deconvolution for the calculation of renal perfusion and glomerular filtration rate (GFR) on the basis of concentration-time curves as measured with perfusion MRI. Six rabbits were scanned dynamically after injection of a gadolinium chelate. Concentration-time curves were generated by manually drawing regions of interest in the aorta and the renal cortex. To remove the dependency on the arterial input function, a regularized structured total least-squares deconvolution algorithm was used to calculate the renal impulse response. This curve was fitted by the sum of two gamma variate functions, corresponding to the passage of the contrast agent in the glomeruli and the proximal convoluted tubules. Tracer kinetics models were applied to these two functions to obtain the renal perfusion and GFR. For comparison, these two parameters were also calculated on the basis of the renal concentration-time curve before deconvolution. The renal perfusion values correlated well (r = 0.9, P = 0.014) with the values calculated by a validated upslope method. The GFR values correlated well (r = 0.9, P = 0.014) with the values obtained from the clearance of (51)Cr-EDTA. A comparison of the values obtained with and without deconvolution demonstrated the necessity of deconvolution.

Inflow effects were studied for T(1)-weighted, fast gradient-echo, contrast-enhanced MRI. This was done on the basis of realistic simulations (e.g., taking slice profiles into account) for unsteady flow. The area under the point spread function (PSF) was used to estimate the flow-related enhancement. A simple analytical model that accurately describes the inflow effects was derived and validated. This model was used to correct the experimental perfusion calibration curves (signal intensity vs. relaxation rate) for inflow effects. Hepatic perfusion measurements, performed on patients, were analyzed in terms of a dual-input, first-order linear model. It was shown that inflow causes incorrect perfusion input functions. The resulting estimated perfusion parameters displayed a systematic error of typically 30-40%. By performing two extra time-resolved flow measurements during the examination, one can correct the input functions.