0552 Estimating Skin Tone from Red and Infrared Light in Existing Pulse Oximeters to Correct Skin Tone-Related Bias

Abstract Introduction Pulse oximetry (SpO₂) systematically overestimates arterial oxygen saturation (SaO₂) in individuals with darker skin pigmentation. This skin tone bias arises because current calibration algorithms do not account for differences in melanin absorption. A potential solution is to develop skin tone-specific calibration curves, but this requires a reliable way to estimateskin tone, preferably using signals already available in commercial oximeters. Because melanin absorbs red light much more strongly than infrared light, we hypothesized that the ratio ofthe red to infrared direct current (DC) signals—representing transmitted rather than pulsatile light—could serve as a quantitative estimate of skin tone. Methods Sixteen participants wereenrolled and categorized using the Fitzpatrick Skin Tone (FST) scale into three groups: light (FST I-II, n=5), medium (FST III-IV, n=4), and dark (FST V-VI, n=7). Using a Texas Instrumentstransmission-mode finger probe with fixed LED drive currents, we recorded 5-second segments of raw red and infrared signals. From these signals, the DC components (representingbaseline light transmission) were extracted, and the ratio of red(DC) to infrared(DC) was calculated for each participant. This ratio was then compared across skin tone categories. Results The red(DC)/infrared(DC) ratio decreased systematically with increasing skin pigmentation. Participants with light skin had ratios >1.0, indicating greater red than infraredtransmission. In the medium group, ratios ranged from 0.6-1.0, consistent with increased red light absorption by melanin. In the dark skin group, most participants had ratios < 0.6, reflectingsubstantial red light attenuation. The pattern demonstrates that the DC ratio distinguishes skin tone categories in a manner consistent with melanin’s wavelength-dependent absorption Conclusion These data suggest that skin tone can be estimated directly from the red and infrared light signals already present in pulse oximeters. Incorporating such a skin-tone metriccould enable skin tone-specific calibration curves and thus substantially reduce or eliminate skin tone bias in pulse oximetry measurements. This work supports the feasibility of a built-in,signal-based method to identify and correct skin tone-related error without requiring additional sensors or external measurement devices. Ongoing analyses will incorporate objectivespectrocolorimetric indices, including the Individual Typology Angle, to further validate this approach. Support (if any)