Evaluation of Pulse Oximetry Accuracy in a Commercial Smartphone and Smartwatch Device During Human Hypoxia Laboratory Testing

TL;DR

This study tests how accurately a Samsung smartphone and smartwatch measure blood oxygen levels in a lab setting with controlled oxygen levels.

Contribution

The study evaluates pulse oximetry accuracy in commercial devices under FDA/ISO standards, including skin tone variations.

Findings

Smartphone and smartwatch RMSD values were 2.6% and 1.8%, meeting FDA/ISO standards.

More testing is needed for individuals with darker skin tones.

Abstract

Background: The US Food and Drug Administration (FDA) and International Organization for Standardization (ISO) clearance standards for the clinical use of smart device pulse oximetry require in-laboratory human hypoxemia testing in healthy human individuals using arterial blood gas analysis. Methods: We evaluated the SpO2 measurements of the Samsung smartphone (Galaxy S9/10) and smartwatch (Galaxy 4) at stable arterial oxygen saturations (SaO2) between 70 and 100% in 24 healthy participants. Testing followed FDA/ISO-stipulated procedures for pulse oximetry performance validation, which include questionnaire estimation of skin tone based on Fitzpatrick estimation of skin types I–VI. During testing, inspired oxygen, nitrogen, and carbon dioxide partial pressures were monitored and adjusted via partial rebreathing circuits to achieve stable target arterial blood oxygen (SaO2) plateaus between 70% and 100%. Arterial blood samples were taken at each plateau, with device SpO2 readings taken at each sample extraction. An ABL-90FLEX blood gas analyzer determined arterial blood sample SaO2. Bias, calculated from device readings minus corresponding arterial blood measurements, was reported as root mean square deviation (RMSD). Results: Combined Participants demographics were: 62.5% female; median age 26 years (range 21–46); and race/ethnicity 16.7% African American, 33.3% Asian, 12.5% multi-ethnic, and 37.5% Caucasian. Fitzpatrick Skin Scale-identified skin tones were: white–fair (I&II), 20.8%; average–light brown (III–IV), 54% and brown–black (V–VI), 25%. There were no adverse events. The RMSD values of SpO2 measurements were: smartphone 2.6% (257 data pairs) and smartwatch 1.8% (247 data pairs). Conclusions: Device SpO2 demonstrated RMSD < 3.0% to SaO2, meeting FDA/ISO clearance standards at the time of study. However, additional testing in persons with darker skin tones is necessary. Smartphones and paired wearables, when cleared for clinical use following revision of FDA clearance standards, may expand access to remote pulse oximetry.

Figures