Neuronal Activation Detection Using Vector Phase Analysis with Dual Threshold Circles: A Functional Near-Infrared Spectroscopy Study

Abstract

In this paper, a new vector phase diagram differentiating the initial decreasing phase (i.e. initial dip) and the delayed hemodynamic response (HR) phase of oxy-hemoglobin changes ( $\Delta$ HbO) of functional near-infrared spectroscopy (fNIRS) is developed. The vector phase diagram displays the trajectories of $\Delta$ HbO and deoxy-hemoglobin changes ( $\Delta$ HbR), as orthogonal components, in the $\Delta$ HbO- $\Delta$ HbR polar coordinates. To determine the occurrence of an initial dip, dual threshold circles (an inner circle from the resting state, an outer circle from the peak values of the initial dip and the main HR) are incorporated into the phase diagram for making decisions. The proposed scheme is then applied to a brain-computer interface scheme, and its performance is evaluated in classifying two finger tapping tasks (right-hand thumb and little finger) from the left motor cortex. Three gamma functions are used to model the initial dip, the main HR, and the undershoot in generating the designed HR function. In classifying two tapping tasks, the signal mean and signal minimum values during 0-2.5 $$ s, as features of initial dip, are used. The linear discriminant analysis was utilized as a classifier. The experimental results show that the active brain locations of the two tasks were quite distinctive ( $p<0.05$ ), and moreover, spatially specific if using the initial dip map at 4 $$ s in comparison to the map of HRs at 14 $$ s. Also, the average classification accuracy was improved from 59% to 74.9% when using the phase diagram of dual threshold circles.

Keywords: Functional near-infrared spectroscopy (fNIRS); brain–computer interface (BCI); designed hemodynamic response function; initial dip; motor cortex; neuronal firing; vector phase analysis.