Revisiting random cylinder models in IVIM: biophysical simulations of realistic microvascular networks reveal asymmetries and heterogeneity

Motivation: Cerebral cortical IVIM signals lack ground-truth knowledge of blood velocities and segments for proper interpretation. Goal(s): To glean microvascular and hemodynamic information impacting the IVIM signal and identify sizes of blood pools in different regimes. Approach: We report the structural and hemodynamic properties from biophysical simulations of blood flow in realistic microvascular reconstructions of mouse cortical gray matter and a synthetic human counterpart. Results: The blood velocities were asymmetric between descending arterioles and ascending venules. Blood does not change direction during the mixing time (Δ) during the human brain IVIM protocol, suggesting a voxel may contain "diffusive" (~20%) and "ballistic" (~40%) blood regimes. Impact: Microvascular architecture and dynamics are not random and both influence the IVIM signal, yet interpretation currently relies on simplified vascular models using random cylinders. We characterized IVIM parameters in realistic microvasculature of the brain, revealing biases absent in simpler models.