A simple permutation-based test of intermodal correspondence.

Area of Science:

• Neuroimaging
• Statistical analysis
• Brain mapping

Background:

• Neuroimaging studies often rely on identifying similarities between brain maps.
• Current statistical methods for assessing these similarities have limitations, including strong and unrealistic assumptions like covariance stationarity.
• Existing methods typically compare group-level brain maps against spatial null models.

Purpose of the Study:

• To propose a novel statistical framework for testing and assessing similarities between brain maps using subject-level data.
• To overcome the limitations of current methods, particularly their reliance on unrealistic statistical assumptions.
• To provide a more flexible and reliable approach for localizing intermodal relationships within specific brain regions.

Main Methods:

• Utilized a classical permutation testing framework with subject-level neuroimaging data.
• Randomly permuted subjects to generate a null distribution of intermodal correspondence statistics.
• Compared observed statistics against the null distribution to estimate p-values.

Main Results:

• The proposed method demonstrated good performance in detecting known relationships between modalities (e.g., cortical thickness and sulcal depth).
• The method was conservative when no association was expected (e.g., cortical thickness and n-back task activation).
• The approach proved flexible and reliable for localizing intermodal relationships within subregions of the brain.

Conclusions:

• The subject-level permutation testing framework offers a more flexible and reliable alternative for analyzing brain map similarities in neuroimaging.
• This method allows for generalizable statistical inference and precise localization of intermodal relationships.
• The approach addresses limitations of traditional methods by avoiding strong statistical assumptions.