Non-parametric combination and related permutation tests for neuroimaging.

Area of Science:

• Neuroimaging analysis
• Statistical inference
• Multivariate statistics

Background:

• Multiple testing is a common challenge in neuroimaging, particularly with diverse data types and hypotheses.
• Existing methods for combining and correcting for multiple tests can be computationally intensive and inflexible.

Purpose of the Study:

• To adapt permutation methods for flexible combination analyses in neuroimaging.
• To develop an efficient, single-phase non-parametric combination (NPC) methodology for joint inference.
• To evaluate and identify optimal combining methods for permutation tests.

Main Methods:

• Synchronized permutations for multiplicity correction across various data types (imaging, non-imaging, different resolutions).
• Modified non-parametric combination (NPC) methodology for single-phase joint inference.
• Evaluation of various combining methods (e.g., Tippett's method) within permutation tests.

Main Results:

• The proposed modified NPC method offers efficient joint inference with reduced computational demands and storage.
• Permutation-based approaches provide flexibility in integrating multimodal and multi-resolution neuroimaging data.
• Tippett's method is identified as effective for linking multiplicity correction and combination.
• The method demonstrates favorable comparisons to classical multivariate tests like MANCOVA.

Conclusions:

• Permutation methods provide a robust and flexible framework for complex neuroimaging analyses involving multiple tests and data types.
• The developed single-phase NPC method enhances computational efficiency for joint inference.
• The findings offer practical guidance for selecting optimal combining strategies in neuroimaging research.