High-resolution MRI Guided Whole Mouse Brain Cell Type Atlas using Deep Learning.

Area of Science:

• Neuroscience
• Bioimaging
• Computational Biology

Background:

• Cell types are fundamental units of brain organization, traditionally studied using single-cell sequencing.
• Single-cell sequencing offers molecular detail but lacks whole-brain resolution.
• Diffusion MRI (dMRI) provides macroscale architectural insights and is sensitive to neurodevelopment and disease.

Purpose of the Study:

• To determine if dMRI can directly predict cell types in the mouse brain.
• To develop a high-resolution, whole-brain cell atlas using integrated imaging techniques.
• To establish an efficient strategy for brain cell atlas generation.

Main Methods:

• Developed a deep learning framework integrating high-resolution dMRI with 3D light-sheet microscopy (LSM).
• Registered dMRI and LSM data to the Allen Mouse Brain Common Coordinate Framework (CCFv3).
• Investigated correlations between dMRI metrics and spatial transcriptomics-derived cell types.

Main Results:

• Generated a whole-brain cell type atlas at 10 micrometer isotropic resolution.
• Demonstrated the capacity of the deep learning framework to predict cell types from dMRI data.
• Established significant correlations between dMRI metrics and cell type distribution.

Conclusions:

• Advanced imaging and deep learning provide an efficient, high-resolution strategy for brain cell atlas generation.
• dMRI holds significant potential for non-invasively probing brain cytoarchitecture and cell type distribution.
• This work advances our understanding of cellular mechanisms within the brain through integrated imaging approaches.