Improving wearable out-of-hospital pulse oximetry estimation with novel real-time analysis of photoplethysmography

Cardiopulmonary emergencies and chronic cardiopulmonary illnesses remain leading causes of morbidity and mortality, largely because events occurring outside medical supervision go unwitnessed. Wearable sensing offers the possibility of detecting acute physiological decline in real-time, yet current commercial systems are constrained by limitations in sensitivity, specificity, validation, mathematical and data processing paradigms for monitoring hypoxic physiological states, and user-adherence barriers. Specifically, SpO₂ - a common indicator of arterial oxygenation - requires a pulse to provide a result, thus failing during low perfusion states or cardiac arrest. This thesis investigates whether acute desaturation can be continuously monitored using photoplethysmography (PPG)-based devices to enable rapid detection of acute cardiopulmonary instability. To address this question, three complementary aims were pursued. A pan-Canadian population survey examined user preferences and willingness to adhere to continuous wearable use for critical illness detection. Responses (n ≈ 400) revealed a strong preference for hand-based form factors and identified age- and activity-dependent differences in adherence behavior, underscoring the need for user-centered design and public education on critical illness risk. Building on these insights, the second aim advanced the analytical foundations of PPG-derived oxygen saturation (SpO₂) estimation. A derivative Beer–Lambert model was developed to infer changes in arterial oxygenation (ΔSpO₂) independent of pulsatile blood flow, introducing a calibration coefficient α that remained generalizable across participants. A multivariate Kalman filter was subsequently implemented to fuse instantaneous SpO₂ and derivative information (dDiff), yielding semi-lagless SpO₂ tracking with markedly improved temporal resolution, tested in a simulated benchtop hypoxemia experiment to establish model accuracy relative to a clinical pulse oximeter. The third aim validated these methods in experimental simulated pulselessness. This arterial-occlusion study demonstrated stability under near-pulseless conditions using a perfusion-index-weighted dynamic Kalman filter. Together, these studies produced a reproducible, open-source pipeline for algorithmic validation in simulated cardiopulmonary illness. Collectively, this work reframes wearable PPG from a passive optical signal into an active estimator of cardiopulmonary deterioration. By integrating user behavior, mathematical modeling, and physiological validation, it establishes a pathway toward real-time, device-agnostic detection of critical illness. The findings have implications for both clinical monitoring and the future of consumer-grade health technologies.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.