Student Carnegie Mellon University Pittsburgh, Pennsylvania, United States
Introduction: NIRS operates on the principle of changes in light intensity due to the absorption of hemoglobin and scattering in the imaged tissue. For this purpose, two or more wavelengths are used within the 650–900 nm range based on the absorption spectra of oxyhemoglobin, deoxyhemoglobin, water, and melanin. There is a significant concern regarding the effect of melanin on NIRS measurements and participant variability across skin tones, as melanin has the potential to reduce light intensity [1], light sensitivity, and signal-to-noise ratio (SNR) [2, 3].
NIRS has widespread applications, including pulse oximetry, which motivates its use in various physiological studies. However, the implications of this research extend to broader hemodynamic measurements, as the potential impact of melanin may affect the accuracy and reliability of both oxyhemoglobin (Δ[HbO]) and deoxygenated hemoglobin (Δ[Hb]) readings. Hemodynamic research is particularly susceptible to these biases, as variations in skin pigmentation can lead to significant discrepancies in data interpretation. In this study, we utilize a paced breathing method to investigate the impact of melanin on monitoring respiration through NIRS measurements. By analyzing the correlation between melanin and the SNR of respiration signals, we aim to better understand and account for the variation in respiratory signal detection introduced by differences in skin pigmentation.
Materials and
Methods: Using a frequency-domain NIRS device (ISS OxiplexTM, Champaign, IL), we measured cerebral hemodynamics from 35 adult healthy participants at 50 Hz sampling frequency. We controlled respiratory signal variation through paced breathing at 0.125 Hz to avoid obstruction in the signal quality from Mayer waves. Five participants were excluded due to excessive motion artifacts or instrumentation issues. We quantified melanin through the melanin index, which was measured with DSM-III. A 3 cm source-detector distance was utilized as it is the most commonly employed distance in NIRS studies, providing a balance between signal quality from the cortical region and depth penetration necessary for reliable cerebral hemodynamics measurement. The change in absorption coefficient (dμ_a ) has been calculated with the Modified Beer-Lambert Law. Oxygenated and deoxygenated hemoglobin have been calculated from dμ_a. The data has been filtered with (respiration frequency – 0.05 Hz) to (respiration frequency + 0.07 Hz) with a 4th-order phase neutral IIR bandpass filter. We then calculated the Spearman correlation between the SNR of the respiration signal for filtered Δ[HbO] and Δ[Hb] with the colorimeter-measured melanin index. To quantify respiration signal-to-noise ratio (SNR), we extracted the power associated with fundamental respiration pulsation and then calculated the ratio with the baseline noise level at 11-19 Hz [4].
The spectral power has been calculated from a power spectrogram utilizing a moving window of 20 seconds, with a 95% overlap to accommodate respiration rate variability. The maximum time-frequency point was then derived using the "tfridge" function of Matlab with a penalty of 0.005.
Results, Conclusions, and Discussions: The SNR of the respiration signal shows a significant negative correlation with the melanin index for both Δ[HbO] and Δ[Hb]. For Δ[HbO], the correlation is moderately negative (r_s= -0.390,p < 0.05) indicating a trend where higher melanin index tends to reduce respiration SNR, (Figure 1a). This trend suggests that melanin may interfere with the accurate detection of oxygenated hemoglobin levels during breathing.
For Δ[Hb], the correlation is strongly negative (r_s= -0.466,p≤0.01), indicating a highly significant trend where an increased melanin index significantly reduces respiration SNR, (Figure 1b). This demonstrates a clearer impact of melanin on the measurement of deoxygenated hemoglobin. The different trends observed for Δ[HbO] and Δ[Hb] highlight the varying effects of skin pigmentation on NIRS measurements. The stronger impact on Δ[Hb] suggests that deoxygenated hemoglobin levels are more susceptible to interference from melanin.
These findings indicate the potential impact of melanin on interpreting physiological signals beyond resting heart rate, emphasizing the necessity to account for melanin concentration in NIRS data analysis for an accurate physiological assessment. All physiological and hemodynamic measurements should be considered to ensure a comprehensive understanding and adjustments for skin pigmentation bias.
In conclusion, the analysis shows further requirements to understand and adjust for melanin's impact on NIRS measurements to ensure accurate interpretation of physiological signals across diverse populations. Given the widespread use of NIRS in monitoring breathing and respiration signals, such as in conditions like hyperventilation and sleep apnea, it is crucial to account for melanin's influence. Future work will focus on developing algorithms to correct signal accuracy, specifically investigating the amplitude, phase difference, respiration shape, and respiration rate of Δ[HbO] and Δ[Hb], to enhance the reliability of NIRS in diverse populations. We will further access these differences to SNR measurements made with pulse oximetry.
.jpg "Alexander Kirages (he/him/his) photo")