An Image-informed Computational Model for Predicting Spatiotemporal Variations in Cerebral Blood Flow

To better understand the control of cerebral blood flow (CBF) and the effect of different CNS pathologies such as seizures on brain hemodynamics, several computational models have been developed. However, many of these models do not recapitulate spatiotemporal CBF variations or pathology-related CBF changes. Recent technological advances in in vivo imaging have enabled us to more accurately account for the microstructural and functional features of the cerebral vasculature in such models and validate their hemodynamic predictions. Therefore, here we present a novel “image-informed model” of temporal CBF changes in healthy and seizure-afflicted murine brains. To calibrate and validate our model, we collected multicontrast image data from four mice under normal and seizure conditions using a multicontrast miniscope developed by our group 1 . Seizures were induced with intraperitoneal injection of Pentylenetetrazole and the intrinsic optical signal and the laser speckle contrast (LSC) channels of the miniscope to estimate segment-specific diameter and relative CBF every 5 seconds. Additionally, we used time-to-peak (TTP) fluorescence derived from injections of FITC-labeled dextran to identify arterioles and venules. Next, we adapted an existing hemodynamics model 2 to incorporate time-varying changes in segment-specific microvascular diameter. This model incorporated Poiseuille’s law and hematocrit-dependent viscosity calculations to iteratively derive nodal pressures and hematocrit distributions using MATLAB ® 2 . Boundary conditions (BC) were assigned using transit time maps, and arteriolar/venular segments identified from TTP data. Inlet boundary pressures were optimized using a diameter-ranked Gaussian distribution, and temporal diameter changes updated with in vivo data to simulate time-varying changes in cerebral hemodynamics (i.e. CBF and CBV). We validated model predictions by comparing CBF velocity distributions with published data and relative CBF changes with LSC-derived measurements. We observed significant correlations between simulated and observed rCBF time-series data for healthy (r ~ 0.75, p < 0.05) and seizure conditions (r ~ 0.5, p < 0.05). Finally, using our validated model we compared the hemodynamics before, during and after a seizure. The predictions showed a decrease in overall vessel diameter, an increase in overall viscosity, and a decrease in CBF during the peak seizure event, compared to the minutes preceding the seizure. We observed the reverse trend for the two minutes after the seizure. Our “image-based-modeling” strategy enabled us to predict time-dependent hemodynamic changes within microvascular networks of the mouse brain by incorporating vessel diameter changes measured in vivo. Our model can be applied to investigate region- and time-dependent hemodynamic variations caused by pathologies such as seizures, strokes and brain tumors. Our model advances the field by employing simulations informed by measurements of individual segments' diameters, and by optimizing BC (i.e. inlet boundary nodes identification and pressure assignment) to limit the number of segments with non-physiological blood flow velocities. In summary, ”image-based models” provide a new platform for investigating temporal alterations in cerebral hemodynamics, enabling the assessment of more physiologically relevant changes in brain function. 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). This work was supported by NIH grant nos. 2R01CA196701-06A1, 5R01CA237597-04 and 5R01DE027957-05 This abstract was presented at the American Physiology Summit 2025 and is only available in HTML format. There is no downloadable file or PDF version. The Physiology editorial board was not involved in the peer review process.