文摘
Systematic errors in quantitative diffusion MRI are primarily due to the interference of the imaging gradients with the gradients used for diffusion sensitization. To compensate for these errors, a correction algorithm, which is validated by phantom measurements, is presented. The imaging results with and without correction are compared with spectroscopic reference data. The crosstalk correction allows one to eliminate the dependence of the evaluated diffusion coefficient on image parameters, thus making it possible to choose different imaging protocols without any compromise in interexperimental comparability of the results. For minimization of the influence of noise on accuracy and reproducibility of the pixelwise-evaluated map of the apparent diffusion coefficient (ADC), a simulation scheme is presented which optimizes gradient factors, the number of differently diffusion-weighted images, the number of signal averagings, and recovery time for given temporal and spatial resolution using the Stejskal-Tanner sequence. Simulations were modeled with previously acquired in vivo data, thus avoiding hypothetical assumptions concerning the actual signal-to-noise and physiological behavior. Inclusion of noise primarily resulted in high ADCs being underestimated. Optimal parameters made this underestimation irrelevant in the physiologically important ADC range. A smaller number of diffusion-weighted images, each with high S/N, yielded much better results than a high number of quite noisy gray-value images.