Automatic Graph-Based Modeling of Brain Microvessels Captured With Two-Photon Microscopy.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Biology

Background:

• Graph models of cerebral vasculature are crucial for understanding brain microphysiology.
• Current automatic graphing methods struggle with complex vascular networks and imaging depth limitations.

Purpose of the Study:

• To develop a fully automatic processing pipeline for generating accurate graph models of microvessels.
• To address challenges in microvessel segmentation and 3D modeling from two-photon microscopy data.

Main Methods:

• A pipeline integrating a fully-convolution neural network for microvessel segmentation.
• A 3D surface model generator and a geometry contraction algorithm for creating single-component graph models.

Main Results:

• Quantitative assessment using NetMets metrics showed low geometric error rates (3.8% false negative, 4.2% false positive) at 60 μm tolerance.
• Topological error rates were also low (6.1% false negative, 4.5% false positive).

Conclusions:

• The proposed automated pipeline efficiently generates accurate and useful graph models of cerebral vasculature.
• This advancement facilitates more reliable studies of brain microphysiology.