Characterizing effects of skin pigmentation and anatomical location in reflective photoplethysmography

Pulse oximetry is a non-invasive method to assess arterial oxygen saturation (SaO2), but recent studies have revealed performance bias with increasing skin tones. To address this, we evaluate the impact of skin pigmentation and anatomical site on reflective oximeter accuracy using a custom photoplethysmography (PPG) device and a medical-grade clinical oximeter for comparison. The custom device included seven VIS-NIR LEDs and photodiodes at four source-detector distances. Data is sampled at 50 Hz with ambient light cancellation and transmitted to a local computer via Bluetooth. Nine anatomical sites are characterized using VIS-NIR reflectance spectroscopy, melanometry, photography, and Monk Skin Tone (MST) classification. Twelve healthy participants (mean age = 30.5 ± 7.2 years, BMI 22.46 ± 1.86), with over 50% in the darkest MST tone bins, underwent a controlled hypoxia protocol (100–70% SaO2) in accordance with FDA guidelines. On conventional medical-grade pulse oximetry clips compared to arterial blood draws of SaO2, a bias of 3.61 % ± 2.77 on dark skin and 1.36 % ± 1.37 on light skin was observed across the saturation range. On custom reflective mode pulse oximeter placed on the on the forehead and the thumb pad, a greater error was observed on the forehead than the thumb which had much less of a range of pigmentation. In all sites, error was observed to increase in more hypoxic ranges. In the future, a melanin index will be derived from raw waveform data to calibrate a skin-tone–inclusive oximetry algorithm. This work supports the design of equitable optical devices across diverse skin tones and body locations.