Multivariate and repeated measures (MRM): A new toolbox for dependent and multimodal group-level neuroimaging data

Area of Science:

• Neuroimaging
• Statistical Analysis
• Multivariate Statistics

Background:

• Conventional neuroimaging group analyses often overlook repeated measures, using multiple between-subject models.
• Existing repeated-measures frameworks in popular software may impose unrealistic covariance assumptions.
• Handling complex designs and multimodal data in group analyses presents significant challenges.

Purpose of the Study:

• To present a novel, efficient multivariate general linear model (GLM) framework for analyzing neuroimaging data with repeated measures.
• To address the limitations of conventional methods in handling complex experimental designs and dependent group data.
• To demonstrate the adaptability of the proposed method for multimodal neuroimaging data integration.

Main Methods:

• Utilizing the multivariate form of the general linear model (GLM).
• Implementation within a new MATLAB toolbox for accessible application.
• Employing permutation methods for robust statistical inference.
• Adapting the framework for multimodal data dependency.

Main Results:

• The multivariate GLM framework provides a simple and efficient solution for analyzing repeated-measures neuroimaging data.
• The method overcomes the drawbacks of conventional approaches, enabling complex comparisons.
• Demonstrated successful adaptation for integrating multimodal imaging data from the same individuals.
• Linear discriminant functions (LDA) are shown for follow-up analysis of multimodal models.

Conclusions:

• The proposed multivariate GLM offers a flexible and powerful approach for group-level neuroimaging analysis with repeated measures and multimodal data.
• This method enhances the analysis of complex neuroimaging study designs, improving statistical rigor.
• The toolbox and framework facilitate advanced integration of multiple imaging techniques for population studies.