High-Density Multi-Distance fNIRS Enhances Detection of Brain Activity during a Word-Color Stroop Task

Abstract

Significance: Functional Near-Infrared Spectroscopy (fNIRS) enables neuroimaging in scenarios where other modalities are less suitable, such as during motion tasks or in low-resource environments. Sparse fNIRS arrays with 30mm channel spacing are widely used but have limited spatial resolution. High-density (HD) arrays with overlapping, multi-distance channels improve sensitivity and localization but increase costs and setup times. A statistical comparison of HD and sparse arrays is needed for evaluating the benefits and trade-offs of HD arrays.

Aim: This study provides a statistical comparison of HD and sparse fNIRS performance to inform array selection in future research.

Approach: We measured prefrontal cortex (PFC) activation during congruent and incongruent Word-Color Stroop (WCS) tasks using both Sparse and HD arrays for 17 healthy adult participants, comparing dorsolateral PFC channel and image results at the group level.

Results: While both arrays detected activation in channel space during incongruent WCS, channel and image space results demonstrated superior localization and sensitivity with the HD array for all WCS.

Conclusions: Sparse channel data may suitably detect activation from cognitively demanding tasks, like incongruent WCS. However, the HD array outperformed Sparse in detecting and localizing brain activity in image space, particularly during lower cognitive load tasks, making them more suitable for neuroimaging applications.

Keywords: Diffuse Optical Tomography; High-Density fNIRS; Pre-Frontal Cortex; Word-Color Stroop; fNIRS.

Figure 1:

For each of the Sparse and High-Density (HD) probe design columns display physical appearance, sensitivity matrices via Monte-Carlo photon path modeling with probe overlay (red dots: sources, blue dots: detectors, pink lines: emphasize ‘grid’ layout of sparse array’s 30mm channels, black/white lines: emphasize ‘hexagonal’ layout of HD array’s 19mm/33mm channels), and Brodmann areas underlying each channel. Sensitivity profile is on a log 10 scale; vertices with values > 0.01 are not masked and not considered part of the relevant sensitivity profile.

Figure 2:

Word-Color Stroop paradigm adapted from Jahani, et al. After instruction and initial rest, 18 blocks of 6x3-s trials each were presented with a jittered inter-block interval (10-15 s). A given block consisted of either all congruent (Easy) trials, or all incongruent (Difficult) trials. The lower-right legend demonstrates accurate user keyboard press for each condition. Order of blocks was randomized for a total of 9 blocks of each condition. Total run time approximates 11min.

Figure 3:

Channels and vertices selected in the region of interest for the Sparse and HD array. On the left panel, black dots mark the center of each channel included in the ROIs. On the right panel, black lines indicate the channels included in the ROIs and the white (unshaded) region of the brain indicates the vertices included. Vertices for both arrays are chosen based on those sensitive to the HD ROI channels.

Figure 4:

Channel space brain response recorded by Sparse and HD arrays during WCS, from Superior view. “HbO Mean”: Group-average hemodynamic response (HbO) for each channel, averaged across 7 to 18 seconds of the blocks for each condition. “T-statistic”: Group-averaged t-statistic of each channel is plotted. Color-scale is grey for absolute values less than t-crit = 2.12 as calculated for 17 subjects.

Figure 5:

Brain and scalp image space brain response recorded by Sparse and HD arrays during WCS, from Anterior view. “HbO Mean”: Group-average hemodynamic response (HbO) for each condition. “T-statistic”: Group-averaged t-statistic of each vertex is plotted. Color-scale is grey for absolute values less than t-crit = 2.12 as calculated for 17 subjects.

Figure 6:

From within the ROIs, group-averaged HbO maximum t-statistics are presented in both channel and brain and scalp image space for each array and WCS conditions. T-critical is 2.12, as calculated for 17 subjects with two-tailed α=0.05. Asterisk indicates p < 0.05 for paired Student’s t-test between arrays (black) and conditions (blue). The subjects’ selected channel or averaged 25 vertices’ concentration time courses are averaged for the timeseries plots. Numerical average, standard error, and paired Student’s t-test values available in Table S3.