Classification of Prefrontal Cortex Activity Based on Functional Near-Infrared Spectroscopy Data upon Olfactory Stimulation

Abstract

The sense of smell is one of the most important organs in humans, and olfactory imaging can detect signals in the anterior orbital frontal lobe. This study assessed olfactory stimuli using support vector machines (SVMs) with signals from functional near-infrared spectroscopy (fNIRS) data obtained from the prefrontal cortex. These data included odor stimuli and air state, which triggered the hemodynamic response function (HRF), determined from variations in oxyhemoglobin (oxyHb) and deoxyhemoglobin (deoxyHb) levels; photoplethysmography (PPG) of two wavelengths (raw optical red and near-infrared data); and the ratios of data from two optical datasets. We adopted three SVM kernel functions (i.e., linear, quadratic, and cubic) to analyze signals and compare their performance with the HRF and PPG signals. The results revealed that oxyHb yielded the most efficient single-signal data with a quadratic kernel function, and a combination of HRF and PPG signals yielded the most efficient multi-signal data with the cubic function. Our results revealed superior SVM analysis of HRFs for classifying odor and air status using fNIRS data during olfaction in humans. Furthermore, the olfactory stimulation can be accurately classified by using quadratic and cubic kernel functions in SVM, even for an individual participant data set.

Keywords: classification; functional near-infrared spectroscopy; hemoglobin response function; machine learning technique; olfaction; prefrontal cortex; support vector machine.

Figure 1

(a) Schematic of the study’s fNIRS system design. (b) Optode placement and channel location on the prefrontal cortex at reference points Fp1 and Fp2, in accordance with the International 10–20 System of placement. This system comprises two sources and two detectors, and each source–detector pair is separated by 2.5 cm.

Figure 2

Schematic representation of the experimental paradigm: A block comprising a 12 s stimulus and a 30 s rest period, along with a 35 s baseline and 20 s termination period.

Figure 3

Flow chart of signal processing using the present machine learning approach to separate odor and air classes through photoplethysmography (PPG-R, PPG-IR, and dbRatio) and hemoglobin response function (including oxyHb and deoxyHb) signal data.

Figure 4

Functional near-infrared spectroscopy measurement of the prefrontal cortex: (a) raw photoplethysmography (PPG) PPG-R and PPG-IR signals, (b) 33 s block data in which the dbRatio waveforms have an odor and air status.

Figure 5

Typical example of raw hemodynamic response function including oxyHb and deoxyHb data, with air and odor as the stimuli, in accordance with the experimental paradigm.

Figure 6

Hemodynamic response of one participant determined through functional near-infrared spectrometry. Subplots show the averaged results for the region of interest from among eight blocks wherein odor (a) and air (b) were used as individual stimuli.

Figure 7

Receiver operating characteristic curves of signals of a dataset with performance of three kernels for an SVM classifier: (a) single signal of oxyHb using the quadratic kernel with an area under the curve of 0.998, and (b) combination of five signals with the optimal area under the curve of 0.72 with the cubic function.

Figure 8

Accuracy comparison for 14 types of data and three kernel functions: linear, quadratic, and cubic.

Figure 9

Comparison of the Z-Score and neglect Z-Score for single (a) and multiple signals (b) and three kernel functions.