Automated administration of medical oxygen using model predictive control incorporating real-time monitoring of breathing parameters

Current automated oxygen delivery systems for spontaneously breathing patients rely solely on peripheral oxygen saturation monitoring and do not account for the influence of breathing parameters on the fraction of inspired oxygen delivered via nasal cannula. This study incorporated breathing parameters that affect the fraction of inspired oxygen, namely, tidal volume and inhalation time, alongside peripheral oxygen saturation monitoring to design a closed-loop control system. A model of respiratory and circulatory systems presented in the literature was used to simulate the peripheral oxygen saturation under patient-specific pathophysiological conditions. A model predictive controller was used to incorporate both breathing parameters and peripheral oxygen saturation measurements into the control system design. For eight different cases, the efficiency of the model predictive controller was compared with that of a tuned proportional-integral-derivative controller, which relied solely on peripheral oxygen saturation monitoring. The model predictive controller achieved a higher efficiency score defined as the difference between time spent in the target range of the peripheral oxygen saturation (between 88% and 92%), and average nasal cannula flow (mL/s) (mean = 0.80 ± 0.07), compared to a proportional-integral-derivative controller (0.73 ± 0.08). The model predictive controller showed the best balance between accurate oxygenation and oxygen consumption with the lowest variability. These findings suggest that a model predictive controller integrating both breathing parameters and peripheral oxygen saturation monitoring offers a reliable and adaptive approach for home-based oxygen therapy in patients with chronic obstructive pulmonary disease, reducing the need for retuning and improving oxygen delivery by balancing oxygenation and oxygen use.