Kavli Affiliate: Terry Jernigan, Anders Dale, Wesley Thompson
| Authors: Chun Chieh Fan, Clare E Palmer, John Iverson, Diliana Pecheva, Oleksandr Frei, Dominic Holland, Wesley K Thompson, Donald Hagler, Terry L Jernigan, Ole Andreassen, Anders M. Dale and Thomas Nichols
| Summary:
The linear mixed effects model (LME) is a versatile modeling approach to deal with correlations among observations. Despite the rising importance of LME due to the complex designs of large-scale longitudinal population neuroimaging studies, LME has seldom been used in whole-brain imaging analyses due to its heavy computational requirements. Here, we introduce a fast and efficient mixed-effects algorithm (FEMA) that makes whole-brain vertexwise, voxelwise and connectome-wise LME analyses possible. In a series of realistic simulations, we demonstrate the equivalency of statistical power and control of type I errors between FEMA and classical LME, while showing orders of magnitude improvement in the computational speed. By applying FEMA on diffusion images and resting state functional connectivity matrices from the Adolescent Brain Cognitive Development StudySM (ABCD) release 4.0 data, we show annualized changes in voxelwise fractional anisotropy (FA) and functional connectomes in early adolescence, highlighting a critical time of maturing connections among cortical and subcortical regions. FEMA enables researchers to quickly examine the relationships between large numbers of neuroimaging metrics and variables of interest while considering complex study designs including repeated measures and family structures.