| dc.description.abstract | Diffusion-weighted imaging (DWI or DW-MRI) is a widely applied and clinically important MRI technique probing the diffusion displacement of water molecules in biological tissue on a micrometer length scale. DWI is however an indirect probe, because the extraction of quantitative diffusion metrics requires modeling of the diffusion signal. A plethora of diffusion models have been adapted trying to accurately characterize and quantify the true biological microstructure. Many of the established diffusion models provide the same type of information, with a varying degree of additional information, and there is no gold standard for when to use a given model of a certain complexity. Two well-established diffusion models of different complexity were investigated in this study; diffusion tensor imaging (DTI) and diffusion kurtosis imaging (DKI). DTI is effective in measuring the dominant direction of water diffusion, however, the model is based upon the assumption that water molecules follow a Gaussian diffusion distribution. Real tissue contains complex cellular structures causing the water molecules to diffuse through highly heterogeneous environments, which leads to a deviation from the Gaussian distribution. DKI is an expansion of the DTI model, including an excess kurtosis term that quantifies the degree of non-Gaussian diffusion. Both models provide information about the standard diffusion parameters; fractional anisotropy (FA), mean diffusion (MD), radial diffusion (RD), axial diffusion (AD). In addition, DKI provides information about mean kurtosis (MK), radial kurtosis (RK) and axial kurtosis (AK). DWI measures diffusion with the use of magnetic gradients, and the precision of the diffusion parameter estimations depends on the signal-to-noise ratio (SNR), number of signal averages (NSA), number of gradient directions, as well as the degree of diffusion-weighting (b-value) used in the image acquisition. In this study, DTI and DKI signals simulating white matter, gray matter and cerebrospinal fluid were generated based on DWI signals extracted from real DW-MRI acquisitions in healthy volunteers. Monte Carlo simulations were performed to investigate the effect of SNR, NSA, number of gradient directions and b-values on the parameter estimations of DTI and DKI, with the aim of optimizing the parameter estimation while keeping the acquisition time at a minimum. The results showed that DTI is more sensitive to noise than DKI in the white matter regions of the brain. In contrast, DKI was more sensitive to noise in the gray matter and cerebrospinal fluid. Increasing the NSA resulted in a general improvement in the parameter estimations for both models. The number of b0-images also had a remarkable influence on the parameter estimations. Using 6 b0-images instead of 1 b0 resulted in a noteworthy increase in precision. The analyses showed that the optimal gradient set for DTI was 6 b0-images and 30 b = 1000 s/mm^2 measurements, while the optimal gradient set for DKI was 6 b0-images, 12 b = 500, 30 b = 1000 and 50 b = 3000 s/mm^2 measurements. | eng |