Poster L13 - Imaging-Informed Computational Model of Murine Cerebral Hemodynamics

Introduction: Cerebral blood flow (CBF) is regulated by various mechanisms such as autoregulation, vasoactive stimuli, neurovascular coupling, and endothelium-dependent responses. Different spatial and temporal computational models have been proposed and applied to enhance our understanding of such mechanisms and predict hemodynamic changes under physiological or pathological conditions. With technological advancements in imaging, computational models have been designed and developed to more accurately account for geometric features of the cerebral vasculature and validate CBF predictions. As a next step in advancing cerebral hemodynamic models, we present a new approach that uses in vivo imaging data to inform the model. Specifically, the 2D/3D geometry of cerebrovascular networks and in vivo measurements of diameter changes over time were used to predict hemodynamics in different murine brain regions.

Materials and

Methods: We employed in vivo data previously acquired from a healthy mouse brain using a multicontrast miniscope developed by our group1. These data included vascular geometry and diameter measurements made every 5s for up to 10 min. We adapted a prior version of the hemodynamic model2, which used geometrical vascular features (node connections, vascular coordinates, vessel length, and diameter) as inputs to incorporate temporal diameter measurements. The brain region's hemodynamic characteristics were calculated based on the 1D formulation of Poiseuille’s law and the hematocrit-dependent viscosity law. Using MATLAB®, we iteratively solved a system of algebraic equations that satisfied mass conservation within vessel nodes to derive nodal pressure and hematocrit distributions. Pressure and hematocrit values for each vascular segment were iterated until the system converged. The base model has been previously derived from experimental data by Pries et al. To assign boundary conditions, we employed the fluorescence intensity map derived by injecting FITC-labeled Dextran in the tail vein. We assigned arteriolar/venular boundary nodes (inputs/outputs) based on the time-to-peak in the fluorescence map. Arteriolar boundary nodes (BNs) were assigned 75 mmHg pressure, and venular nodes were assigned 15 mmHg pressure and scaled by segment diameter. We predicted hemodynamic changes over time by updating segment diameter using the in vivo data.

Results, Conclusions, and Discussions: The cerebrovascular network contained 553 segments, corresponding to a field of view of 300x300 μm^2 (Fig. 1A). The high miniscope resolution (6 μm) allowed us to measure diameters ranging from 6 to 120 μm. From the FITC data, we identified a distribution of 28 arteriolar (inputs) and 15 venular (outputs) boundary nodes (Fig. 1B). Implementing a 10-minute simulation of diameter changes, we obtained an average blood flow velocity of ~8.1±9.95 mm/s. Fig. 1C illustrates the evolution of blood flow from 5s-10min. The predicted velocity, pressure, viscosity, hematocrit, and shear distributions were obtained by combining data from four simulation time points (5s, 1 min, 5 min, and 10 min) and are shown in Figs. 1D-H. Our model predictions were within the range reported in previous works for the murine brain2. Working with 2D visualizations to extract physiologically accurate hemodynamic characteristics was challenging. Determining boundary inputs and outputs was the main source of uncertainty that affected the model predictions and defined the pressure and flow distribution throughout the vascular tree. Additionally, assumptions were required to determine the connectivity of overlapping segments, which can also affect the model predictions. Although 94% of our segment velocity predictions fell within the physiological range, a few outliers with velocities > 50 mm/s were also identified. Our simulations indicated that the distribution of inputs/outputs influenced the number of outliers, and future efforts will be directed at addressing this issue. In summary, our imaging-informed hemodynamic model is novel for predicting dynamic hemodynamic changes based on experimental measurements of in vivo diameter changes and 2D visualizations of the vascular network geometry in the murine brain. Future approaches will include improving boundary conditions to avoid outliers and prediction validation using 3D vascular network data. Such imaging-informed models have many potential applications, such as investigating CBF dynamics in CNS pathologies ranging from stroke to brain cancer.

References:
1. Senarathna, J. et al. A miniature multi-contrast microscope for functional imaging in freely behaving animals. Nat Commun 10, 99 (2019).
2. Bhargava, A. et al. VascuViz: a multimodality and multiscale imaging and visualization pipeline for vascular systems biology. Nat Methods 19, 242–254 (2022).

Acknowledgements (Optional): This work was supported by NIH/NCI grant nos. 2R01CA196701-06A1, 5R01CA237597-04 and and 5R01DE027957-05.