Multinomial inference on distributed responses in SPM.

Area of Science:

• Neuroimaging
• Statistical Analysis
• Brain Mapping

Background:

• Traditional statistical inference in neuroimaging focuses on local features (e.g., peak height) within statistical parametric maps (SPMs).
• Limited methods exist for inferring the spatial distribution of topological features like clusters in SPMs, especially with complex noise and geometry.

Purpose of the Study:

• To develop and present novel statistical methods for inference on the spatial distribution of topological features in SPMs.
• To introduce a Bayesian approach for detecting distributed response patterns in SPMs characterized by anisotropic, non-stationary noise and arbitrary geometry.

Main Methods:

• Proposed statistical methods for spatial inference on topological features in SPMs.
• Developed a Bayesian framework to detect experimentally-induced distributed responses.
• Extended the framework for fixed- and random-effects analyses at within- and between-subject levels.

Main Results:

• Successfully applied a Bayesian approach to characterize the spatial distribution of topological features in SPMs.
• Demonstrated the method's capability to handle complex noise and arbitrary geometries.
• Illustrated the application in characterizing language anatomy at various functional segregation scales.

Conclusions:

• The proposed statistical methods offer a robust framework for spatial inference on topological features in SPMs.
• The Bayesian approach effectively detects distributed responses, accommodating complex data characteristics.
• The method provides a powerful tool for neuroimaging research, particularly in understanding functional organization like language networks.