Solving the where problem and quantifying geometric variation in neuroanatomy using generative diffeomorphic mapping.

Area of Science:

• Neuroscience
• Bioinformatics
• Computational Biology

Background:

• Mapping neuronal cell types across whole vertebrate brains is a key focus in neuroscience.
• Existing workflows struggle with multimodal signals, missing data, and quantifying individual variation.
• Integrating diverse molecular and anatomical datasets into a common atlas presents significant challenges.

Purpose of the Study:

• To develop a quantitative workflow for unifying multi-modal whole-brain images within an atlas framework.
• To address limitations in existing methods for mapping complex neuroscientific datasets.
• To enable large-scale integration of diverse datasets for advancing neuroscience research.

Main Methods:

• Implementation of a generative model for data likelihood given atlas transforms.
• Utilized a maximum a posteriori estimation framework for data integration.
• Developed a method for composing mappings across chains of datasets and computing geometric quantification metrics.

Main Results:

• Quantified cell density and geometric fluctuations across various datasets (MRI, fMOST, histology, snRNAseq).
• Demonstrated that individual variation in datasets is often greater than differences introduced by tissue processing.
• Validated the workflow using mouse brain datasets.

Conclusions:

• Established a robust, quantitative workflow for multi-modal whole-brain data integration in an atlas.
• The developed framework successfully unifies diverse imaging and molecular data.
• Facilitates large-scale data integration, crucial for advancing comprehensive neuroscience understanding.