Quantifying the Impact of Hair and Skin Characteristics on Signal Quality with Practical Recommendations for Improvement

Abstract

Functional Near-Infrared Spectroscopy (fNIRS) holds transformative potential for research and clinical applications in neuroscience due to its non-invasive nature and adaptability to real-world settings. However, despite its promise, fNIRS signal quality is sensitive to individual differences in biophysical factors such as hair and skin characteristics, which can significantly impact the absorption and scattering of near-infrared light. If not properly addressed, these factors risk biasing fNIRS research by disproportionately affecting signal quality across diverse populations. Our results quantify the impact of various hair properties, skin pigmentation as well as head size, sex and age on signal quality, providing quantitative guidance for future hardware advances and methodological standards to help overcome these critical barriers to inclusivity in fNIRS studies. We provide actionable guidelines for fNIRS researchers, including a suggested metadata table and recommendations for cap and optode configurations, hair management techniques, and strategies to optimize data collection across varied participants. This research paves the way for the development of more inclusive fNIRS technologies, fostering broader applicability and improved interpretability of neuroimaging data in diverse populations.

Keywords: Functional Near-Infrared Spectroscopy; hair characteristics; inclusivity; signal quality; skin pigmentation.

Figure 1.

Optode array design. Anterior view (A), right view (B) and posterior view (C). Optode array design with sources (red dots), detectors (blue dots), and channels (yellow lines). Standard 10-20 locations are marked on the head for reference. Probe sensitivity is visualized on the brain surface, with values represented in log10 units.

Figure 2.

Violin plots illustrating the relationship between the hair characteristics and the Corrected Signal Mean on the Side of the Head. The data are divided into four equal-width bins based on the hair characteristics, with different colors representing each bin. Violin plots show the distribution of the fNIRS signal data for each bin, with boxplot elements overlayed to indicate quartiles, medians, and potential outliers. Significant differences between to first and the last bins are marked with asterisks (*, p < 0.01 after Benjamini-Hochberg correction). A green line indicates a statistically significant increase, and a red line indicates a statistically significant decrease.

Figure 3.

Violin plots illustrating the relationship between the hair characteristics and the Corrected Signal Mean on the Back of the Head. The data are divided into four equal-width bins based on the hair characteristics, with different colors representing each bin. Violin plots show the distribution of the fNIRS signal data for each bin, with boxplot elements overlayed to indicate quartiles, medians, and potential outliers. Significant differences between bins are marked with asterisks (*, p < 0.01 after Benjamini-Hochberg correction). A green line indicates a statistically significant increase, and a red line indicates a statistically significant decrease.

Figure 4.

Violin plots illustrating the relationship between the hair characteristics and the Corrected Signal Mean on the Side and Back of the Head. Each category is represented with a different color. Violin plots show the distribution of the fNIRS signal data for each bin, with boxplot elements overlayed to indicate quartiles, medians, and potential outliers. Significant differences between the first and last bins are marked with asterisks (*, p < 0.01 after Benjamini-Hochberg correction). A green line indicates a statistically significant increase, and a red line indicates a statistically significant decrease.

Figure 5.

Violin plots of the Corrected Signal Mean on the Forehead, Side of the Head and Back of the Head versus the skin pigmentation. The data are divided into four equal-width bins based on the skin pigmentation, with different colors representing each bin. Violin plots show the distribution of the fNIRS signal data for each bin, with boxplot elements overlayed to indicate quartiles, medians, and potential outliers. Significant differences between the first and the last bins are marked with asterisks (*, p < 0.01 after Benjamini-Hochberg correction). A green line indicates a statistically significant increase, and a red line indicates a statistically significant decrease.

Figure 6.

Violin plots of the Corrected Signal Mean on the Forehead, on the Side of the Head and on the Back of the Head versus head circumference (top) and age (bottom). The data are divided into four equal-width bins based on the head circumference and age, with different colors representing each bin. Violin plots show the distribution of the fNIRS signal data for each bin, with boxplot elements overlayed to indicate quartiles, medians, and potential outliers. Significant differences between the first and last bins are marked with asterisks (*, p < 0.01 after Benjamini-Hochberg correction). A green line indicates a statistically significant increase, and a red line indicates a statistically significant decrease.

Figure 7.

Violin plots of the Corrected Signal Mean on the Forehead, Side of the Head and Back of the Head versus sex. Violin plots show the distribution of the fNIRS signal data for each bin, with boxplot elements overlayed to indicate quartiles, medians, and potential outliers. Significant differences between the first and last bins are marked with asterisks (*, p < 0.01 after Benjamini-Hochberg correction), indicating statistical significance assessed using the Mann-Whitney U Test with Benjamini-Hochberg correction for multiple comparisons. A green line indicates a statistically significant increase, and a red line indicates a statistically significant decrease.

Figure 8.

Cohen’s f² effect sizes from the multiple linear regression analysis predicting fNIRS-signal quality on the Forehead, Side of the Head and Back of the Head. Effect sizes were interpreted according to Cohen’s (1988) guidelines, where f² ≥ 0.02 (green horizontal line) indicates a small effect, f² ≥ 0.15 a medium effect (blue horizontal line), and f² ≥ 0.35 a large effect.

Figure 9.

Cohen’s f² effect sizes from the multiple linear regression analysis predicting GLM output ΔHbO₂ and t-statistics for a ball-squeezing task. Effect sizes were interpreted according to Cohen’s (1988) guidelines, where f² ≥ 0.02 (green horizontal line) indicates a small effect, f² ≥ 0.15 a medium effect (blue horizontal line), and f² ≥ 0.35 a large effect.

Figure 10.

A) Scatter plot of Uncorrected Signal Mean in individual channels across all participants versus the Combined Hair-Skin Metric. The y-axis is in logarithmic scale. Grey dashed horizontal lines represent noise equivalent power levels (NEP/√(Hz)), beneath which any signal drowns in measurement noise. The measurement system’s NEP is 52fW/√(Hz) at an equivalent signal input noise level of 6.8*10⁻⁶ and a detector area of 9.6 mm². The red area highlights all channels that must be omitted if an SNR signal quality threshold of two orders of magnitude (20dB) above the NEP is applied. Example: In a system with an NEP of 1pW, with a 20dB threshold, no channel below the vertical line at 100pW would pass the quality constraint. This is further quantified in subplot B) For each binned group of Combined Hair-Skin Metric, the plot shows the percentage of channels that pass a 20dB signal quality threshold relative to system sensitivity on the x-axis. Example: For the third group in purple color (i.e. Combined Hair-Skin Metric ∈ [0.45-0.6]), 98% of all channels would pass the quality threshold for a system with sensitivity of 10fW/sqrt(Hz), but only 70% of all channels would pass the threshold for a system with sensitivity of 1pW/sqrt(Hz).

Figure 11.

NinjaCap