MesoNet allows automated scaling and segmentation of mouse mesoscale cortical maps using machine learning

Area of Science:

• Neuroscience
• Computational Biology
• Image Analysis

Background:

• Understanding brain function necessitates analyzing cortical activity across spatial scales.
• Accurate anatomical atlasing of functional brain data is labor-intensive and requires expertise.
• Current methods for mesoscale cortical image analysis can be challenging for high-throughput studies.

Purpose of the Study:

• To develop an automated, machine learning-based pipeline for registration and segmentation of mouse mesoscale cortical images.
• To facilitate quantitative analysis of brain activity by aligning functional data with anatomical atlases.
• To provide a user-friendly toolbox for high-throughput analysis of brain imaging data.

Main Methods:

• Developed a deep learning model to identify nine cortical landmarks from single fluorescent images.
• Adapted a fully convolutional network for precise brain boundary delimitation.
• Integrated anatomical alignment with three functional alignment approaches using sensory maps and activity motifs.

Main Results:

• MesoNet accurately identifies cortical landmarks and brain boundaries in mouse mesoscale images.
• The pipeline provides robust anatomical alignment with the Allen Mouse Brain Atlas.
• The system is designed as a user-friendly, Python-based toolbox for efficient data analysis.

Conclusions:

• MesoNet offers an automated and robust solution for quantitative analysis of mouse cortical images.
• This methodology significantly reduces the labor and expertise required for brain data analysis.
• MesoNet enhances the capacity for high-throughput analysis in neuroscience research.