Machine learning analysis of whole mouse brain vasculature

Area of Science:

• Neuroscience
• Bioimaging
• Computational Biology

Background:

• Tissue clearing enables 3D imaging of biological specimens without sectioning.
• Analyzing large, 3D imaging datasets of brain vasculature presents scalability and reliability challenges.

Purpose of the Study:

• To develop a deep learning-based framework for quantifying and analyzing brain vasculature in cleared mouse brains.
• To enable scalable and unbiased analysis of angioarchitecture.
• To gain biological insights into brain vascular function.

Main Methods:

• Developed Vessel Segmentation & Analysis Pipeline (VesSAP), a deep learning framework utilizing a convolutional neural network (CNN) with transfer learning.
• Achieved human-level accuracy for vessel segmentation.
• Applied VesSAP to analyze whole C57BL/6J, CD1, and BALB/c mouse brains at micrometer resolution after registering to the Allen mouse brain atlas.

Main Results:

• VesSAP enables accurate, scalable quantification of brain angioarchitecture.
• Identified secondary intracranial collateral vascularization in CD1 mice.
• Observed reduced brainstem vascularization compared to the cerebrum across analyzed strains.

Conclusions:

• VesSAP provides an unbiased and scalable method for analyzing cleared mouse brain vasculature.
• The framework facilitates biological discoveries regarding vascular function and inter-strain differences.