Diffusion-MRI-Based Estimation of Cortical Architecture via Machine Learning (DECAM) in Primate Brains.

Area of Science:

• Neuroscience
• Biophysics
• Medical Imaging

Background:

• Cerebral cortical cytoarchitecture is crucial for brain function and is altered in aging and disease.
• Conventional methods for studying cytoarchitecture are invasive, limiting whole-brain analysis.
• Diffusion MRI (dMRI) offers noninvasive potential but current models have limitations in accurately quantifying cortical architecture.

Purpose of the Study:

• To introduce Diffusion-MRI-based Estimation of Cortical Architecture using Machine-learning (DECAM), a novel framework for noninvasive mapping of primate brain cytoarchitecture.
• To overcome limitations of current dMRI models for accurate quantification of cortical architecture.
• To enable direct, whole-brain mapping of soma density in primates.

Main Methods:

• Developed DECAM, a data-driven, machine-learning framework integrating high-resolution multi-shell dMRI and histological data from non-human primate brains.
• Optimized the deep learning framework using a novel best response constraint.
• Utilized cortical label vectors to address dMRI-histology misregistration in complex primate brain morphology.

Main Results:

• DECAM accurately and directly maps heterogeneous, whole-brain soma density in primates.
• Generated high-fidelity, reproducible whole-brain soma density maps validated by histology.
• Demonstrated the generalizability of the DECAM framework.

Conclusions:

• DECAM provides a noninvasive method for virtual histology, advancing translational research.
• The framework can be extended to estimate other neuropathological measures like neurite density.
• DECAM holds potential for application in human brain imaging for disease research.