Simplified Anisotropic IVIM Using Spherical Means and an Application in the Placenta

To reduce the complexity associated with anisotropic IVIM modeling, we derived two simplified diffusion MRI models that map the apparent diffusivity, relative proportions, and anisotropy of perfusion- and diffusion-related signal components and demonstrated them in placenta scans.

We introduce two spherical mean anisotropic intravoxel incoherent motion (IVIM) models. The models consist of constrained diffusion tensor compartments, namely "sticks" or "zeppelins," and we consider the spherical mean of the signal for these compartments. We apply these spherical mean anisotropic IVIM models to placenta MRI scans from 14 control participants to produce perfusion fraction, perfusion-related pseudo-diffusivity, diffusion-related diffusivity, perfusion-related fractional anisotropy, and diffusion-related fractional anisotropy maps. We compare our approach to the standard IVIM model and calculate the Bayesian information criterion voxelwise to determine which model best explains the data.

Spherical mean anisotropic IVIM output maps are consistent with standard IVIM, while additionally accounting for and mapping anisotropy of perfusion- and diffusion-related signal components. Spherical mean anisotropic IVIM models provide a better explanation of the data, as indicated by a lower Bayesian information criterion compared to standard IVIM in 10%-20% of voxels, while retaining sensitivity to the same underlying tissue and perfusion characteristics in the remaining voxels.

Our spherical mean anisotropic IVIM approach can disentangle perfusion- and diffusion-related anisotropy without explicitly estimating directionality. Our approach simplifies the calculation of biomarkers reflecting microcirculatory and microstructural changes in anisotropic tissue by employing models of reduced complexity compared to previous methods.