A Functional-Near-Infrared-Spectroscopy-Based Evaluation of Cognitive Lagging From Prefrontal Hemodynamics

Abstract

Functional near infrared spectroscopy (fNIRS) devices capture the variability in oxygenated and deoxygenated blood flow in the cortical layers of the brain during execution of cognitive tasks. The present letter employs fNIRS device to study the cognitive lagging in working memory (WM) performance based on visuo-spatial forward span number search. The unsupervised learning-based clustering yields three distinct clusters of hemodynamic loads during the task performance. From these three clusters, we estimate cognitive lagging by computing the performance score per unit time and assigned three cognitive load classes: low, moderate, and high. Moving toward the classification approach of these three cognitive load classes, ensemble learning classifier produces higher classification accuracy, which reaches a maximum of 91.66%. The trend of shifting cognitive load from low toward high with performance score is observed from estimated Pearson's cross-correlation using the medoid points of cognitive load clusters and associated performance scores. The visualization of dynamic changes in cognitive load (low, moderate, and high) in temporal span of WM performance is obtained from the voxel plot approach, which advocates that regional deactivation of orbitofrontal cortex and augmented hemodynamic load in the dorsolateral prefrontal cortex has a possible relation with this cognitive lagging. In the view of experimental outcomes, the fNIRS-sensor-based measuring of cognitive load could be a future assessment tool for cognitive failure in higher task demand.