Dependence of IVIM-DWI estimations on acquisition parameters

Dissertação para obtenção do Grau de Mestre em Engenharia Biomédica

Nonalcoholic fatty liver disease (NAFLD) is one of the most common chronic liver conditions in the Western world. It is normally associated with health disorders such as obesity, diabetes and hypertriglyceridemia. The gold standard for the diagnosis and staging of NAFLD is liver biopsy, which is highly invasive and is associated with high morbidity and inherent sampling error. Intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI) is able to distinguish between true molecular diffusion and the pseudo-random motion of water molecules inside micro vessels. IVIM-DWI has emerged in the recent years as a possible alternative to probe tissue changes in NAFLD. However few studies have addressed the problem of the dependence of IVIM-DWI parameters on pulse sequence parameters. In order to study this dependence, two studies were carried-out: 1) A simulation study, where we studied the influence of acquisition parameters on the error and bias associated with IVIM-DWI parameters; 2) In-vivo study in order to test the performance of the b-value sequences derived from the simulation studies. Results showed that the parameter which is more affected by the acquisition parameters is D*. Furthermore, it was also shown that the higher the number of b-values used to sample the data, the better the estimation of IVIM-DWI parameters is. However, after a certain number of points and for low SNRs, the effect of noise in extra b-values counteracts the effect of having more data points. It was also shown that the b-value sequence that is used to sample the data greatly influences IVIM-DWI estimations. We concluded that the conventionally used b-value sequence does not provide optimum IVIM-DWI estimations. Furthermore, results show that different weights should be attributed to each IVIM-DWI parameter in order to obtain a better performance of the optimized b-value sequence. Also, it was seen that the influence of T2 relation effects should be accounted for in the IVIM-DWI model. Lastly, our study showed that in the presence of steatosis, the value D* significantly decreased while D only slightly decreased. However, the differences between patients with steatosis and healthy controls were extremely influenced by the number of b-values used, leading to different diagnosis depending on the number of b-values used in the acquisition.