| Summary: | Functional near infrared spectroscopy (fNIRS) is a neuroimaging techonology that enables investigators to indirectly monitor brain activity in vivo through relative changes in the concentration of oxygenated and deoxygenated hemoglobin. One of the key features of fNIRS is its superior temporal resolution, with dense measurements over very short periods of time (100ms increments). Unfortunately, most statistical analysis approaches in the existing literature have not fully utilized the high temporal resolution of fNIRS. For example, many analysis procedures are based on linearity assumptions that only extract partial information, thereby neglecting the overall dynamic trends in fNIRS trajectories. The main goal of this article is to assess the ability of a functional data analysis approach for detecting significant differences in hemodynamic responses recorded by fNIRS. Children with and without specific language impairment wore two, 3*5 fNIRS caps situated over the bilateral parasylvian areas as they completed a language comprehension task. Functional data analysis was used to decompose the high dimensional hemodynamic curves into the mean function and a few eigenfunctions to represent the overall trend and variation structures over time. Compared to the most popular general linear model, we did not assume any parametric structure and let the data speak for itself. This analysis identified significant differences between the case and control groups in the oxygenated hemodynamic mean trends in the right inferior frontal cortex and left inferior posterior parietal cortex brain regions. We also detected significant group differences in the deoxygenated hemodynamic mean trends in the right inferior posterior parietal cortex and left temporal parietal junction brain region. These findings, using dramatically different approaches, experimental designs, data sets, and foci, were consistent with several other reports, confirming group differences in the importance of these two areas for syntax comprehension. The proposed functional data analysis was consistent with the temporal characteristics of fNIRS, thus providing an alternative methodology for fNIRS analyses.
|
|---|
