Articles techniques et présentations

Arterial oxygen partial pressure, PaO₂, is a critical physiological marker for respiratory assessment, hypoxia detection, and critical care management, yet its measurement typically requires invasive arterial blood gas sampling [1], [2]. Transcutaneous oxygen monitoring offers a promising route toward noninvasive oxygen assessment; however, conventional systems often rely on local heating to arterialize the skin and enhance oxygen diffusion [3], [4]. This heat dependence can limit long-term wearable use and may increase the risk of skin irritation or thermal injury, particularly in vulnerable populations [3], [4]. Non-heated wearable PtcO₂ sensors provide a safer and more practical alternative, but their clinical utility is limited by the complex, patient-specific relationship between blood oxygenation and oxygen measured at the skin surface. Variations in skin architecture, microvascular perfusion, tissue metabolism, temperature, oxygen transport properties, and sensor–tissue interaction can cause the same PaO₂ level to produce different PtcO₂ readings across individuals [4], [5].

We present a personalized three-dimensional multiphysics digital twin developed in COMSOL Multiphysics to model and interpret wearable transcutaneous oxygen measurements. The model couples capillary blood flow, oxygen diffusion through multilayered skin, nonlinear hemoglobin–oxygen dissociation, tissue oxygen consumption, and oxygen transport to the sensor interface. This framework captures how arterial oxygen changes are filtered by the microvasculature, surrounding tissue, and skin layers before appearing as a measurable PtcO₂ signal [5], [6].

The primary innovation is the transition from a standard forward oxygen-transport simulation to a patient-specific digital twin for model-informed sensor interpretation. By adjusting key physiological and sensor-related parameters, the model can represent different virtual patient profiles and quantify how individual variability shifts the PaO₂–PtcO₂ relationship. Representative hypoxia and reduced-perfusion simulations will be used to demonstrate how virtual patients with similar PaO₂ levels can produce distinct PtcO₂ responses due to differences in local tissue and vascular properties. The workflow combines forward simulation, sensitivity analysis, and inverse estimation to identify the dominant contributors to the measured sensor response. This enables a physics-based approach for estimating the likely range of underlying PaO₂ values from wearable PtcO₂ measurements while accounting for individual physiological variability [5], [7].

Beyond physiological insight, the digital twin serves as a medical device design and optimization platform. It can be used to assess sensor placement, required sensitivity, membrane properties, response time, calibration protocols, and robustness under hypoxia or reduced-perfusion conditions. By linking wearable sensor output to patient-specific oxygen transport physiology, this work demonstrates the potential of multiphysics simulation to accelerate the development of personalized, noninvasive respiratory monitoring technologies [7], [8].