A Quantitative Approach For Estimating Intravoxel Incoherent Motion (IVIM) Model Parameters In Diffusion Weighted Magnetic Resonance Imaging



Journal Title

Journal ISSN

Volume Title



Diffusion weighted magnetic resonance imaging (DW-MRI) has found numerous clinical applications like tumor classification, recognizing acute ischemic stroke, etc. Tissue perfusion is traditionally evaluated by monitoring MRI signal changes following the administration of contrast agents. However, if tissue diffusion is modeled as consisting of a vascular compartment and a tissue compartment with relatively lower diffusivity (compared to a vascular compartment), then tissue perfusion information can be obtained using DW-MRI without contrast administration. This model is called intra-voxel incoherent motion (IVIM) model of biologic tissue. Despite its promise, the IVIM model has not gained widespread clinical acceptance for three main reasons: (a) IVIM derived perfusion metrics are noisy and lack precision, (b) the lack of in-vitro models that can mimic wide physiological conditions encountered in-vivo, and (c) the dependency of perfusion related IVIM (pr-IVIM) model parameters on acquisition parameters. In this dissertation, I propose a set of solutions to address these limitations. First, I propose and demonstrate via numerical simulations a new approach - analytical segmented (AS) approach- to improve the robustness of estimation of pr-IVIM model parameters. Second, I demonstrate the design and implementation of an in-vitro phantom model with the flexibility to adjust the pr-IVIM model parameters over a wide range of physiological conditions. Third, I propose and demonstrate an extended AS approach - called AS-T2 method- capable of extracting pr-IVIM model parameters with little sensitivity to acquisition parameters such as echo time (TE) (AS-T2 method), and provides a means to extract the T2 relaxation parameter of the fluid compartment (hitherto not described). Fourth, the accuracy, precision, and bias of my proposed analysis approaches are rigorously evaluated using in-silico and in-vitro models.

Finally, preliminary in-vivo results in human brain parenchyma show that pr-IVIM model parameters extracted using the proposed approaches have a lower coefficient of variation than conventional methods and are relatively insensitive to changes in TE.

In sum, the algorithmic approaches and the in-vitro phantom model described in this dissertation can be used to standardize results from data acquired across different commercial MRI platforms. More extensive clinical trials are necessary to confirm these findings.



Diffusion weighted imaging (DWI), Intravoxel incoherent motion (IVIM) model, IVIM phantom, IVIM model analysis algorithms, T2 correction for IVIM model