Nested and robust modeling techniques for fNIRS data with demographics and experiment related factors in n-back task

Abstract

Functional near-infrared spectroscopy (fNIRS) signals are used to measure relative changes in oxyhemoglobin (HbO) and deoxyhemoglobin (HbR) concentrations. Brain response studies constitute multilevel or nested datasets formed by different parts of the brain of individuals and multidimensional datasets. The changes in brain activities under specific stimuli are investigated with the help of statistical analysis. However, these studies ignore the dependence structure between the repeated measures of the same subject, which may cause inaccurate or incomplete findings. In this study, we adopt an advanced statistical method into HbO data controlling for variability within repeated measures of each subject while testing and measuring the degrees of the statistical significance between-subject factors and explanatory variables. The changes in HbO are investigated through a linear mixed model, taking experimental and demographic variables into account with open access neuroscience data. The channels nested within subjects are considered random to capture the differences among individuals. Our findings reveal that n-back conditions and mean response times of the subjects have statistically significant associations with mean HbO.