Multimodal Branched Transport Infers Anatomically Aligned Brain Reaction Maps

Area of Science:

• Neuroscience
• Computational Biology
• Systems Neuroscience

Background:

• Understanding how external stimuli generate brain-wide responses is crucial but challenging.
• Current models often rely on predefined network structures and do not infer the underlying propagation architecture.
• The transformation of localized stimulation into distributed neural activity patterns remains a key question.

Purpose of the Study:

• To infer the brain's propagation architecture from activity data.
• To develop a model that estimates stimulation and reaction measures and anatomical transport costs.
• To investigate the trade-off between geometric efficiency and dynamical controllability in neural signal routing.

Main Methods:

• Combined task-related blood-oxygen-level-dependent (BOLD) responses, source-reconstructed electrophysiology, and tractography-derived anisotropy.
• Estimated stimulation and reaction measures and defined an anatomical transport cost.
• Employed variational optimization to infer a branched propagation architecture and a stochastic graph-induced dynamical model.

Main Results:

• Multimodal data successfully generated anatomically aligned brain reaction maps.
• Anisotropic costs significantly altered routing backbones compared to isotropic models.
• Hybrid geometric-dynamical optimization identified non-trivial rank reversals in branching regimes.

Conclusions:

• The study infers a novel branched transport architecture for neural signal propagation.
• Anatomical constraints and multimodal data integration are vital for accurate brain-wide modeling.
• The findings offer new insights into brain connectivity and information flow dynamics.