Replicated Gibbs Point Process to Combine Functional Brain Alterations Patterns in Acute Sleep Deprivation Using Coordinate Based Meta-Analysis

Neuroimaging studies on the brain activation patterns in acute sleep deprivation (SD) are conducted with different conditions of experiment and have reported inconsistent results. Thus replicability of meta-analysis results might be questionable if the effect of confounders be ignored. Our goal is firstly, to improve replicability of our meta-analysis by considering the heterogeneity between the results of SD studies in framework of mixed effect modeling. Secondly, involving interactions between reported foci in our analysis.Method Conditional intensity function (CIF) is a conventional tool for modeling in Gibbs point process. CIF defined as the average number of foci at a desired pixel, given the rest of the reported foci. If acquired functional brain alterations patterns from each study be considered as a realization of brain activity at the population level, one could account unobserved variability due to different studies as a random effect and use mixed effect model to estimate CIF. Geyer saturation process with variational Bayes technique is used to compute posterior estimate of parameters [1].Results Our investigations are motivated by data from [2]. The data have been retrieved from 31 publications on SD. Main covariates are: average of age and hours of SD per each study. Our results provide evidence that hours of SD has an important effect on the function of SD. Thus it’s expected that total number of foci for a study with above 24 hours of SD be 2.12 times more than if it was under 24 (p-value=0.016). Figure1 shows predicted CIF as measure of activity pattern for hours of SD. The average of age has no significant effect (p-value=0.21).Conclusions In this work, a replicated model was used to separate different sources of variability in activity pattern of SD. Also we find that CIF give a better capture of brain activity than estimated intensity which is a conventional measure in model based meta-analysis.