Accuracy of gamma-variate fits to concentration-time curves from dynamic susceptibility-contrast enhanced MRI: influence of time resolution, maximal signal drop and signal-to-noise

Abstract

Concentration-time curves derived from dynamic susceptibility-contrast enhanced magnetic resonance imaging are widely used to calculate cerebrovascular parameters. To exclude effects of recirculation, a non-linear regression method is used to fit a gamma-variate function to the concentration-time course. In previous studies the errors arising from the fitting procedure have not been quantified. In a computer simulation we investigate the uncertainties of parameters calculated from the fitted gamma-variate function, exploring the dependencies on signal-to-noise (SNR), time resolution (delta t), and maximal signal drop (MSD). Our study was performed to give a framework on how to design MR-sequences and choose contrast media and their application in order to yield concentration-time curves which allow a reliable performance of the gamma-variate fitting procedure. We recorded 396 concentration-time curves from regions of interest of 40 patients. The gamma-variate fitting procedure was applied to these curves resulting in 396 parameter sets. Ideal concentration-time curves as gamma-variate functions were generated from these sets with a given delta t, MSD, and SNR. Recirculation effect was simulated. Then the gamma-variate fitting was performed again. From ideal and simulated gamma-variate function the area and the normalized first moment were calculated. The uncertainties of the values calculated from the simulated curve relating to the values of the original one were determined. Increase of SNR decreases the involved errors. With SNR values of 100 and more there is only minor influence of delta t and MSD and the fitted curve approximates the original data very well. Smaller values of SNR lead to a stronger influence of delta t and MSD and a higher number of fitting failures. With increasing delta t the uncertainties also increase. Intermediate values of MSD (30% to 70%) yield the smallest errors while increasing or decreasing MSD yields an increase of uncertainty. To achieve low uncertainties in the calculation of cerebrovascular parameters from gamma-variate fits, delta t of the imaging sequence and MSD must be considered. This is more important the lower SNR is. The shown dependencies should be taken into account when choosing MR sequence parameters and application of contrast media.