Spatiotemporal linear mixed effects modeling for the mass-univariate analysis of longitudinal neuroimage data

Abstract

We present an extension of the Linear Mixed Effects (LME) modeling approach to be applied to the mass-univariate analysis of longitudinal neuroimaging (LNI) data. The proposed method, called spatiotemporal LME or ST-LME, builds on the flexible LME framework and exploits the spatial structure in image data. We instantiated ST-LME for the analysis of cortical surface measurements (e.g. thickness) computed by FreeSurfer, a widely-used brain Magnetic Resonance Image (MRI) analysis software package. We validate the proposed ST-LME method and provide a quantitative and objective empirical comparison with two popular alternative methods, using two brain MRI datasets obtained from the Alzheimer’s disease neuroimaging initiative (ADNI) and Open Access Series of Imaging Studies (OASIS). Our experiments revealed that ST-LME offers a dramatic gain in statistical power and repeatability of findings, while providing good control of the false positive rate.

Increasing Body Mass Index at Midlife is Associated with Increased Cortical Thinning in Alzheimer’s Disease-Vulnerable Regions

“…LME is a standard analysis approach for longitudinal data and correctly models the mean and covariance structure of repeated measures within participants and across assessments. The temporal covariance structure was assumed to be shared across vertices within homogenous regions of interest [18].…”

Instantiated mixed effects modeling of Alzheimer's disease markers

“…modeling a time-shift as random effect [22], using predefined sequences of marker abnormality [53] or iteratively from long-term progression curves and subject-specific linear random effects [12]). Moreover, methods that rely on clinical variables for grouping in their analysis can be said to also implicitly use such grouping as a rough form of temporal alignment [4,5].…”

“…Detailed studies into early state longitudinal Alzheimer's disease marker trajectory dynamics, using data-driven methods, have the potential to aid the effort in the development of measures that can accurately and robustly quantify indications of the disease, even before its presymtomatic and preclinal stages. Previously, hypothetical [23,24] and experimental models [13,14,50,10,20,7,25,41,21,4,5,1,12,15,53] of disease progression based on Alzheimer's disease markers, such as cerebrospinal fluid (CSF), imaging and cognitive markers have been proposed.…”

“…Recently, several approaches that regard the disease progression trajectory as a continuous process have been developed [13,14,50,10,20,7,25,41,21,4,5,1,12]. Many of the proposed frameworks rely on modeling cognitive scores such as the mini mental state examination (MMSE), Alzheimer's disease assessment scale (ADAS), clinical dementia rating sum of boxes (CDRSB), or Rey's audio visual learning test (RAVLT) among others, as surrogate measures of disease progression.…”

“…To this end, Ito and Hutmacher [21] proposed the use of a longitudinal non-linear mixed effects model to predict the CDRSB time course, which is used to estimate the median time for clinically worsening (defined by several thresholds in change of CDRSB) and to determine clinical endpoints to use in therapeutic trials. Bernal-Rusiel et al [4,5] proposed spatio-temporal linear mixed effects modeling for mass-univariate analysis of longitudinal neuroimaging data. Adak et al [1] calculated rates of cognitive decline using hierarchical mixed effects models in which each subject is assumed to have a linear trajectory of cognitive decline.…”