Device Technologies and Biomedical Robotics
Vivian Krause (she/her/hers)
Undergraduate Student
Elon University
Elon, North Carolina, United States
Pulse oximeters are widely accessible health monitoring devices that estimate oxygen saturation (SpO2) non-invasively and continuously through optical measurements. These devices are widely used for at-home health monitoring and clinical applications, however, many studies dating back to the 1990’s have reported biases in pulse oximetry based on numerous factors, including skin tone. In 2020, an FDA briefing warned about inaccuracies in these devices, coming after a large-scale multi-cohort study, which found that the occurrence of occult hypoxemia in darker skin patients was three times that in light-skinned patients. Occult hypoxemia refers to patients with arterial oxygen saturation below 88%, despite SpO2 readings between 92% and 96%. The impact of this variance in accuracy was magnified during the COVID-19 pandemic, as many relied on these devices for health monitoring. In this work, I present a dual-sensor system for pulse oximetry that measures skin tone and corrects for skin pigmentation in blood oxygen readings.
Results and Discussion:
A prototype was successfully developed that identifies user skin tone and utilizes the Monte Carlo simulation data for corrected oxygen saturation readings. Through a combination of raw data captured by the sensor, user-reported skin types, and image processing through MATLAB, three classes (A: light pigmentation, B: moderate pigmentation, & C: dark pigmentation) were identified and assigned ranges used to calibrate the prototype. Considerations of user-friendliness and functionality were used to design and 3D print a pulse oximeter clip and sensor mount to ensure consistent conditions for readings. MCMATLAB was used to simulate reflectance pulse oximetry readings in a multilayer skin model and this data was used to develop calibration curves for each tested melanin concentration, and correction curves for each designated skin class.
Conclusions:
Here, I present a prototyped dual-sensor system for pulse oximetry that reads user skin tone and takes skin tone into account when computing SpO2. This design combines calibration techniques from experimental data for the skin-tone classification, and theoretical models for the SpO2 correction curves. Photon absorption through multilayer biological tissue was modeled with Monte Carlo simulations in MATLAB. This study demonstrates the potential for a device that provides an objective measurement of user skin tone to correct pulse oximetry readings. However, further field testing will be necessary to advance this device in practice.