FEMA: Fast and efficient mixed-effects algorithm for large sample whole-brain imaging data.

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biostatistics

Background:

• Linear mixed-effects models (LME) are crucial for analyzing dependent observations in complex neuroimaging studies.
• High computational demands of LME have limited its application in large-scale, whole-brain analyses.
• Existing methods struggle with the complexity of modern neuroimaging datasets.

Purpose of the Study:

• Introduce a fast and efficient mixed-effects algorithm (FEMA) for whole-brain LME analyses.
• Enable vertex-wise, voxel-wise, and connectome-wide LME on large neuroimaging datasets.
• Facilitate advanced investigations of neuroimaging metrics considering complex study designs.

Main Methods:

• Developed a novel fast and efficient mixed-effects algorithm (FEMA).
• Validated FEMA using extensive simulations comparing its estimates to standard maximum likelihood estimates.
• Applied FEMA to longitudinal resting-state fMRI data from the Adolescent Brain Cognitive Development Study.

Main Results:

• FEMA provides equivalent fixed-effect estimates to standard methods but with significantly improved computational speed.
• Analysis of adolescent brain data revealed distinct spatial patterns of cortical thickness and functional connectivity changes.
• Identified critical periods of brain maturation during early adolescence.

Conclusions:

• FEMA makes whole-brain LME analyses feasible for large-scale neuroimaging datasets.
• The algorithm efficiently handles complex designs, including repeated measures and family structures.
• FEMA enables deeper insights into brain development and relationships between neuroimaging metrics and variables of interest.