• Login
• Admin

Kuppur Narayana Swamy, Suvvi (2025) Peripheral oxygen saturation measurement across diverse skin tones. PhD thesis, University of Nottingham.

PDF (Thesis - as examined) - Repository staff only until 1 September 2027. Subsequently available to Anyone - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader Available under Licence Creative Commons Attribution. Download (16MB)

Abstract

Pulse oximeters (PO)s have been used in hospitals for over fifty years, yet challenges in peripheral oxygen saturation (SpO2) estimation due to skin tone variations were historically considered insignificant and largely overlooked. However, the COVID-19 pandemic shifted this perspective, as numerous retrospective and prospective studies revealed that POs frequently overestimated SpO2 in individuals with darker skin tones compared to white individuals. In view of this, this thesis investigated the impact of skin tone on pulse oximetry through a combination of bench testing, Monte Carlo simulations, clinical trials, and novel algorithm development.

First, a bench-top optical simulator was developed to evaluate commercial transmission-mode pulse oximeters. By generating controlled PPG-like signals and mimicking different levels of melanin absorption and noise conditions with the help of light attenuators, the system allowed reproducible testing across saturation ranges. When tested across different saturation ranges, all POs showed no bias in SpO2 readings, regardless of whether equal or unequal attenuation filtering occurred at red and infrared (IR) wavelengths. This was because the filters mimicked the effect of melanin in a simplified manner, resulting in similar AC and DC light attenuation at both red and IR wavelengths. Consequently, this effect rendered the ‘ratio of ratios’ R calculation and SpO2 estimation unaltered.

Following this, to address the limitations of a simplified simulation testbench, a Monte Carlo-based semi-infinite four-layer finger model was employed. This was aimed at exploring the effect of complex light-tissue interaction on SpO2 readings, particularly epidermal absorption and scattering. Simulations were performed in both transmission mode and reflection mode. In transmission mode, SpO2 overestimation was reproduced when the finger was modelled as a cylindrical geometry, but not with a rectangular tissue slab. The rectangular geometry resulted in equal attenuation of AC and DC components at both wavelengths. However, the cylindrical geometry introduced unequal interactions at the epidermal interface, that led to relatively greater reduction in AC and perfusion index (PI) at the red wavelength, which in turn caused SpO2 overestimation in darker tissue model. In reflection mode, significant variations in SpO2–R curves were observed between light and dark skin tones. Additionally, as the source-detector separation distance increased, the differences in SpO2–R curves between skin tones diminished, suggesting that a larger separation distance may reduce the impact of skin pigmentation on SpO2 readings in reflection mode.

To improve SpO2 estimation and minimize error due to motion artefacts, a novel Six Median R extraction algorithm was developed by the author. This algorithm was evaluated in two clinical trials conducted in 2017 on newborns (N=13), and in 2023 on healthy adults (N=13). In the 2017 study, SpO2 measured on the newborn’s forehead using a reflectance mode sensor was significantly higher (p<0.005) compared to a peripherally sited PO, suggesting preferential perfusion of the core region immediately after birth. In the 2023 study, two reflectance mode commercial sensors (Surepulse Medical, VS(2020) and VSP(2023)) were assessed on the forehead of 13 healthy adults who underwent induced hypoxia. The sensors were tested under the BS-EN-ISO-80601-2-61:2019 standard to generate an SpO2 calibration curve with over 200 points while meeting demographic requirements. Cohort level SaO2-R calibration curves were generated, achieving an RMSE of 4%. With increase pigmentation, the slope of the calibration curves increased, indicating SpO2 overestimation in darker skin tones.

Finally, building upon these insights novel full-spectrum PPG analysis method was developed. The feasibility of this approach was first investigated using Monte Carlo simulated data from a cylindrical finger model. This method applied the modified Beer Lambert law to systolic and diastolic states to obtain the change in absorbance (difference) spectrum. The predicted SpO2 was obtained by optimizing to difference spectrum using the known absorption coefficients of oxyhaemoglobin (HbO2), deoxyhaemoglobin (HHb) and melanin. The method demonstrated significant improvement in SpO2 estimation in both light and dark finger models compared to the conventional ‘ratio of ratios’ approach.

Overall, this work advances understanding of how skin tone affects SpO2 measurement, introduces novel algorithms, and provides insights that may inform future device design and regulatory evaluation of pulse oximeters.

Item Type:	Thesis (University of Nottingham only) (PhD)
Supervisors:	Hayes-Gill, Barrie Morgan, Steve Correia, Ricardo Liu, Chong
Keywords:	Oxygen saturation, skin tone, Pulse oximetry, Photoplethysmography, Occult hypoxemia, Monte Carlo, Sensors
Subjects:	Q Science > QP Physiology
Faculties/Schools:	UK Campuses > Faculty of Engineering > Department of Electrical and Electronic Engineering
Item ID:	82394
Depositing User:	Kuppur Narayana Swamy, Suvvi
Date Deposited:	31 Dec 2025 04:40
Last Modified:	31 Dec 2025 04:40
URI:	https://eprints.nottingham.ac.uk/id/eprint/82394

Actions (Archive Staff Only)

Edit View