Equity-Driven Sensing System for Measuring Skin Tone–Calibrated Peripheral Blood Oxygen Saturation (OptoBeat): Development, Design, and Evaluation Study (Preprint)

Abstract

BACKGROUND Many commodity pulse oximeters are insufficiently calibrated for patients of darker skin. We demonstrate a quantitative measurement of this disparity in SpO2 measurement with a controlled experimental set up using synthetic skin. To mitigate this, we present OptoBeat, an ultra-low-cost smartphone based optical sensing system that captures SpO2 and heart rate while calibrating for differences in skin tone. Our sensing system can be constructed from commodity plastics, fiber-optic cable and three clips that can be 3D printed for approximately $1, or cheaply manufactured at scale. In our experiments, we demonstrate the efficacy of the OptoBeat system, which can measure SpO2 levels within 1% accuracy of the ground truth (an FDA approved pulse oximeter) in SpO2 levels as low as 75%. OBJECTIVE The objective of this work is to test the following hypothesis and implement an ultra-low-cost smartphone adaptor to measure SpO2. • H1: Skin tone has a significant effect on pulse oximeter measurements. • H2: Pulse oximeter error based on skin tone can be corrected if skin tone is known. • H3: SpO2 can be measured with a smartphone camera using the screen as a light source. METHODS We used three tones of synthetic skin (Syndaver), with the same optical and chemical properties as human skin, to conduct all ex-vivo experiments. A skin tone scale was printed out and placed in the images captured by a mobile phone to calibrate and serve as a ground truth. To achieve a wide range of SpO2 measurements, we used sheep blood (Hardy Diagnostics) that was reoxygenated in a pressure chamber and pulsed through synthetic arteries with a peristaltic pump system. Custom optical clips coupled with fiberoptic cables focus the light from a smartphone screen through the analyte into the phone’s camera. SpO2 measurements are captured by pulsing the screen red and blue. RESULTS Skin tones were accurately classified as being type 2, 3, and 5 on the Fitzpatrick scale using the Euclidian distance of the captured RGB values. Traditional pulse oximeter measurements showed significant differences between skin tones in both AC and DC measurements. The standard deviations in the ratio of IR/red were 0.4184% for type 5, 0.2484% for type 3, and 0.2536% for type 2. Results show a significant difference between the three skin tones as shown in the results of an ANOVA test: 5997 Degrees of Freedom, F score of 3.1170e+05, and p < 0.001. Using our system, SpO2 measurements between 98-75% blood oxygen saturation were reliably captured in an ex-vivo experiment and are accurate to within 1% of ground the truth. In the human subject’s validation experiment, SpO2 measurements were accurate to within 0.5% of ground truth and pulse rate measurements were accurate within 2% of the ground truth. CONCLUSIONS Skin tone has a significant effect on SpO2 measurements using standard commodity hardware. This can be corrected by normalizing for variations in skin tone using an RGB image and reference scale. Leveraging existing smartphone hardware, we classify skin tone, measure SpO2, and normalize the measurements. To do this, we designed OptoBeat, an ultra-low-cost optical system