Characterizing the influence of skin pigmentation on pulse oximetry

Arjun Putcha, Kevin Schichlein, Thang Nguyen-Tien, Brayden Davis, Rachel Choffin, S. Malik, Aastha Sharma, Grace Sosa, Sara Saffari, Joseph G. Ribaudo, Jessica Guidici, Richard M. Lampman, J. Eian, Michael R. Kosorok, Leonid Shmuylovich, Mitchell A. Pet, Wubin BaiBiophotonics 2025 · DOI

TL;DR

This paper investigates why pulse oximeters often overestimate blood oxygen levels in individuals with darker skin tones. By using a live porcine model with naturally pigmented and non-pigmented skin patches, the researchers demonstrate that skin pigmentation significantly attenuates the sensitivity of red light absorption to oxygen saturation changes, leading to reduced pulse oximeter accuracy in darker skin.

What Problem Does This Paper Address?

Pulse oximeters are a ubiquitous, non-invasive tool for measuring blood oxygen saturation (SpO2), crucial for clinical decision-making. However, a well-documented problem is their tendency to overestimate SpO2 in patients with darker skin tones, a bias that worsens at lower oxygen levels. While skin pigmentation has been suspected as a cause, the precise mechanisms and quantitative impact have remained largely unelucidated. This paper aims to bridge that gap by experimentally characterizing how skin pigmentation directly influences the intermediate calculations of pulse oximetry.

Methods

The researchers developed a mathematical framework to model photoplethysmography (PPG) signals in both time and frequency domains. They then applied this framework to data acquired from a live porcine model. This model was chosen for its naturally occurring patches of pigmented and non-pigmented skin, allowing for direct comparison.

Key methodological aspects include:

  • Porcine Model: Utilizing a live porcine model with distinct pigmented and non-pigmented skin patches.
  • Simultaneous Measurement: Recording PPG data from both pigmented and non-pigmented skin simultaneously during induced deoxygenation.
  • Gold Standard Comparison: Collecting arterial blood gas (ABG) samples concurrently to serve as a ground truth for oxygen saturation.
  • Digital Signal Processing (DSP): A robust pipeline was employed to filter raw PPG data, extract pulsatile (AC) and baseline (DC) components, and analyze them in both time and frequency domains. This included median filtering, EW-MAF, Chebyshev Type II Low Pass Filter, FFT/iFFT for component extraction, and periodic moving average filters.
  • Signal Quality Indices (SQIs): Various metrics (Pulsatility Index, relative power spectral density, spectral kurtosis, spectral skew, autocorrelation) were used to select valid analysis windows and ensure data reliability.
  • Statistical Analysis: Paired t-tests and regression t-tests were used to compare the slopes and intercepts of regression lines between pigmented and non-pigmented skin across different parameters (modulation ratio, AC/DC ratios) and SaO2 levels.

Datasets

The study did not use pre-existing datasets for training or evaluation. Instead, it generated a novel dataset through experimental measurements on a live porcine model. This dataset consists of simultaneous PPG recordings (red and infrared light) and arterial blood gas (ABG) measurements from both pigmented and non-pigmented skin patches across various induced oxygen saturation levels.

Key Insights and Ideas

  • Melanin's Attenuating Effect: The core insight is that melanin in pigmented skin disproportionately attenuates red light absorption, which is a critical component for pulse oximetry calculations. This attenuation directly impacts the sensitivity of the oximeter to changes in oxygen saturation.
  • Red Light vs. NIR Light: The study highlights that the impact of pigmentation is more pronounced in the red light spectrum compared to near-infrared (NIR) light. This is consistent with the known optical properties of melanin.
  • Reduced Sensitivity, Not Just Bias: The findings suggest that the issue isn't just a simple offset bias, but a fundamental reduction in the sensitivity of the pulse oximeter's red light signal to changes in blood oxygen levels in pigmented skin.
  • Validation of Clinical Observations: The experimental results strongly support and provide a mechanistic explanation for the long-observed clinical phenomenon of pulse oximeter overestimation in patients of color.
  • Mathematical Framework Utility: The developed mathematical framework for PPG signal analysis proved effective in characterizing the subtle differences caused by pigmentation.

Key Results

  • Reduced Slope in Modulation Ratio: A statistically significant difference was found in the slope of the modulation ratio (R) versus SaO2 regression lines between pigmented and non-pigmented skin, with pigmented skin showing a lower slope. This indicates reduced sensitivity to SaO2 changes.
  • AC Ratio Sensitivity: The AC ratio (pulsatile component) also showed significantly different slopes and intercepts between pigmented and non-pigmented skin, further supporting the reduced sensitivity in pigmented skin.
  • Red Pulsatile Index Attenuation: The red pulsatile index (PI) was found to be significantly more attenuated in pigmented skin, especially at lower SaO2 levels (hypoxia), while the NIR pulsatile index remained largely unaffected.
  • Lower Agreement in Pigmented Skin: The adjusted R-squared values between pigmented and non-pigmented skin showed a statistically significant lower degree of agreement, indicating less reliable SpO2 estimates in pigmented skin.

Limitations and Future Work

  • Limited Testing on Vascularization: The authors acknowledge that their testing on the impact of vascularization on pulse oximeter accuracy was limited. They suggest that future work should explore this further, noting that they found better agreement between core and peripheral locations in non-pigmented skin compared to pigmented skin.
  • Reflectance-Mode Oximeter: The study exclusively used a reflectance-mode pulse oximeter. Further testing with transmission-mode pulse oximeters is recommended to confirm generalizability.
  • Sample Size: The authors acknowledge the limited sample size of their dataset and the need for further testing to replicate their findings.
  • LED Brightness: Keeping LED brightness constant across different skin pigmentations might have impacted signal quality in pigmented skin, potentially affecting interpretability. However, they note that signal quality metrics were used to justify data inclusion.
  • Future Focus: The authors suggest that future research addressing pulse oximeter bias should focus on how light-tissue interactions, particularly in the red region, are uniquely affected by skin pigmentation. They also propose that further work could investigate how to develop targeted methodologies to mitigate this bias.

Summary updated

Abstract

SignificancePulse oximeters are frequently used in healthcare settings to noninvasively estimate blood oxygen levels in patients, in turn influencing clinical decision-making. However, numerous studies have found pulse oximeters consistently overestimate oxygen levels in patients of color, with this overestimation increasing as oxygen levels decline. Although some reports postulate skin pigmentation may be the source of bias, to our knowledge, the precise impact of skin pigmentation on pulse oximetry has not been elucidated.AimOur objective is to experimentally characterize the impact of skin pigmentation on pulse oximeters across a broad range of oxygen saturations and determine whether skin pigmentation can explain the observed pulse oximeter bias in patients of color.ResultsUsing a live porcine model, we experimentally demonstrate the impact of skin pigmentation on pulse oximeter calculation. In particular, we show that skin pigmentation attenuates the sensitivity of red-light absorption to changes in oxygen saturation and show that the pulse oximeter output is consequently attenuated in pigmented skin.ConclusionThe systematic animal experiments confirm the confounding impact of skin pigmentation on intermediate calculations of pulse oximetry, and validate existing clinical observations of pulse oximeter bias in patients of color.
BibTeX
@inproceedings{putcha2025characterizing,
  author = {Arjun Putcha and Kevin Schichlein and Thang Nguyen-Tien and Brayden Davis and Rachel Choffin and S. Malik and Aastha Sharma and Grace Sosa and Sara Saffari and Joseph G. Ribaudo and Jessica Guidici and Richard M. Lampman and J. Eian and Michael R. Kosorok and Leonid Shmuylovich and Mitchell A. Pet and Wubin Bai},
  title = {{Characterizing the influence of skin pigmentation on pulse oximetry}},
  booktitle = {Biophotonics},
  year = {2025},
  doi = {10.1117/1.bios.2.3.032506},
}
Figure