Spatial-extent inference for testing variance components in reliability and heritability studies

Area of Science:

• Neuroimaging
• Statistical genetics
• Brain imaging analysis

Background:

• Existing neuroimaging methods for variance component testing, crucial for heritability and reliability estimation, are limited by the General Linear Model (GLM) and suffer from low statistical power.
• Methodological and computational challenges hinder the development of powerful statistical tests for variance components in neuroimaging data.

Purpose of the Study:

• To develop a fast and powerful statistical test for variance components in neuroimaging data.
• To address the limitations of existing methods in detecting narrow-sense heritability and test-retest reliability.
• To improve statistical power in neuroimaging analyses of genetic and reliability components.

Main Methods:

• Proposed CLEAN-V (CLEAN for testing Variance components), a novel statistical test for variance components.
• Modeled global spatial dependence structure of imaging data.
• Employed data-adaptive pooling of neighborhood information for locally powerful statistics.
• Utilized permutations for family-wise error rate (FWER) control in multiple comparisons.

Main Results:

• CLEAN-V demonstrated superior performance in detecting test-retest reliability and narrow-sense heritability compared to existing methods.
• Achieved significantly improved statistical power in analyses of Human Connectome Project task-fMRI data and simulations.
• Detected areas of significance aligned with functional magnetic resonance imaging (fMRI) activation maps.
• Showcased computational efficiency, indicating practical utility.

Conclusions:

• CLEAN-V offers a significant advancement in statistical testing for variance components in neuroimaging.
• The method provides enhanced power for detecting heritability and reliability, crucial for understanding brain function and individual differences.
• CLEAN-V's computational efficiency and availability as an R package facilitate its widespread application in neuroimaging research.