Measuring Oxygen Saturation With Smartphone Cameras Using Convolutional Neural Networks

Area of Science:

• Biomedical Engineering
• Mobile Health Technology
• Artificial Intelligence in Healthcare

Background:

• Arterial oxygen saturation (SpO2) is a critical indicator of blood oxygen levels, essential for cellular function.
• Traditional SpO2 measurement relies on pulse oximeters, but accessibility and cost can be limitations.
• Emerging research explores using smartphone cameras for non-invasive SpO2 estimation.

Purpose of the Study:

• To develop and evaluate a smartphone-based method for measuring arterial oxygen saturation (SpO2).
• To enhance SpO2 measurement accuracy by mitigating motion artifacts using advanced algorithms.
• To compare the performance of the proposed smartphone method against a standard medical pulse oximeter.

Main Methods:

• Utilized convolutional neural networks (CNNs) for SpO2 estimation from smartphone camera data.
• Implemented specific preprocessing techniques to reduce motion artifacts during measurement.
• Conducted a breath-holding study with 39 participants using two different smartphone models for data collection.

Main Results:

• The proposed smartphone-based SpO2 measurement system demonstrated significantly lower mean absolute error (2.02%) compared to a medical pulse oximeter.
• The system's performance was evaluated across two distinct mobile phone models.
• Comparison with the widely used ratio-of-ratios model showed superior accuracy of the developed method.

Conclusions:

• Smartphone cameras, enhanced with CNNs and robust preprocessing, offer a viable and accurate tool for SpO2 monitoring.
• This technology presents a potential low-cost, accessible alternative for SpO2 assessment in various settings.
• Further validation and development could integrate this method into widespread mobile health applications.