SU-D-303-05: Optimal Protocols and Analysis Strategies for Intravoxel Incoherent Motion MRI

Abstract

Purpose: Intravoxel Incoherent Motion (IVIM) MRI can saperate perfusion and diffusion components of the tissues. Previous studies showed that non-Gaussian diffusion effect may influence the quantifications of perfusion. This study aimed to propose optimal protocols and analysis strategies for the perfusion measurement of normal brain tissues. Methods: An Non-Gaussian IVIM model was used to generate IVIM signals. There were three conditions of b-values: (1) 0–500 (2) 0–1000 (3) 0–2500. SNR of 250 was set. For analysis, the Gaussian and the Non-Gaussian models were used for both full-fitting and asymptotic methods. The mean errors and coefficient of variation were calculated. The MR imaging were applied by using a 3.0T scanner with a 12-channel phase array head coil. The diffusion images ware obtained with the following parameters: b values (0, 5, 10, 20, 30, 50, 70, 100, 200, 400, 600, 800, 1000, 1500, 2000, 2500 s/mm2), repetition time/echo time, 2,500/67 milliseconds; flip angle, 90 degrees; pixel size, 3.4375 x 3.4375 mm2 ; matrix, 64x 64; slice thickness,5 mm; 22 slices, gap= 6 mm; bandwidth, 1594 Hz. Results: Compared with the full-fitting methods, the asymptotic methods provided better precision for IVIM analysis. However, the asymptotic methods were sensitive to the weighting of perfusion component. To minimize the effect of perfusion component, the minimum of the b-value set for asymptotic methods should be larger than 500 and 767 for 5% and 1% residual perfusion component, respectively. In addition, the Gaussian models were sensitive to the non-Gaussian effects. For a general normal kurtosis (K=0.8), the Gaussian model methods were not valid with large b-values. However, the Non-Gaussian methods were sensitive to SNR. The experiment showed similar results with the simulation. Conclusion: The asymptotic Non-Gaussian methods is recommanded for IVIM analysis if appropriate SNR was available. The b-values should be set as 0–2500.