Analysis and prediction of hematocrit in microvascular networks

Hematocrit, defined as the volume percentage of red blood cells in blood, is an important indicator of human health status because it involves the capability of blood to deliver oxygen. The present work aims to analyze and predict the hematocrit distribution in microvascular networks from the direct simulations of cellular blood flow. The networks are geometrically complex, composed of bifurcations, confluences and winding vessels, while the cells exhibit multiple dynamic behaviors, including deformation and aggregation. Hence, our simulations can not only capture the behaviors of each cell in a single vessel, but also characterize the hematocrit of the cell population in a complex network. The results showed that the hematocrit is spatially and temporally heterogeneous in the microvascular network, which is largely attributed to the existence of bifurcations. The more complex the network, the more pronounced the heterogeneity, and it varies steadily over time in wide vessels but periodically fluctuates in narrow vessels. Based on a certain number of data samples, we predicted the hematocrit distribution in a microvascular network, by using the deep learning technique, instead of carrying out the direct simulation. This provides a fast and efficient way to measure the hematocrit distribution in a microvascular network. After that, we examined the effects of other cell types on the hematocrit, which are quite different from the case with the red blood cells only, due to the vascular blockage caused by the cell clusters. In such a case, it becomes very difficult to predict the hematocrit distribution by the deep learning technique, and its heterogeneity becomes more obvious.

Automated summary

This study aims to analyze and predict the hematocrit distribution in microvascular networks using direct simulations of cellular blood flow. The results showed that the hematocrit is spatially and temporally heterogeneous in the microvascular network, with a significant effect from bifurcations. Additionally, deep learning techniques can provide a fast and efficient way to measure the hematocrit distribution in a microvascular network.