A deep learning pipeline for mapping in situ network-level neurovascular coupling in multi-photon fluorescence...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Biology

Background:

• Functional hyperemia, the adjustment of brain blood flow to neuronal activity, is crucial for healthy brain function.
• Previous studies focused on tissue or individual vessels, leaving vascular network-level coordination poorly understood.
• Understanding microvascular network dynamics is essential for comprehending brain function and disease.

Purpose of the Study:

• To develop and validate a deep learning pipeline for automated reconstruction and quantification of microvascular network geometric changes.
• To analyze network-wide vascular responses to neuronal activation in a mouse model.
• To investigate the relationship between neuronal activity, vessel diameter changes, and network efficiency.

Main Methods:

• Developed a deep learning pipeline using two-photon fluorescence microscopy images of cerebral microcirculation.
• Automated reconstruction and quantification of geometric changes in hundreds of interconnected microvessels.
• Applied graph theory-based network analysis to assess changes in network properties like assortativity and efficiency.

Main Results:

• Observed network-wide vessel radius changes (dilations and constrictions) dependent on photostimulation intensity.
• Identified significant heterogeneity in vascular radius adjustments within vessels, linked to contractile cell distribution.
• Demonstrated a substantial increase in network assortativity and a modest increase in capillary network efficiency with increased photostimulation.

Conclusions:

• Individual vessel interrogation is insufficient to predict network-level blood flow modulation.
• The developed pipeline enables dynamic tracking of microvascular network geometry and its relation to neuronal activity.
• This approach can map network-level flow impairments in experimental disease models.