Assessment of spontaneous breathing during pressure controlled ventilation with superimposed spontaneous breathing using respiratory flow signal analysis.

Integrating spontaneous breathing into mechanical ventilation (MV) can speed up liberation from it and reduce its invasiveness. On the other hand, inadequate and asynchronous spontaneous breathing has the potential to aggravate lung injury. During use of airway-pressure-release-ventilation (APRV), the assisted breaths are difficult to measure. We developed an algorithm to differentiate the breaths in a setting of lung injury in spontaneously breathing ewes. We hypothesized that differentiation of breaths into spontaneous, mechanical and assisted is feasible using a specially developed for this purpose algorithm. Ventilation parameters were recorded by software that integrated ventilator output variables. The flow signal, measured by the EVITA® XL (Lübeck, Germany), was measured every 2 ms by a custom Java-based computerized algorithm (Breath-Sep). By integrating the flow signal, tidal volume (V) of each breath was calculated. By using the flow curve the algorithm separated the different breaths and numbered them for each time point. Breaths were separated into mechanical, assisted and spontaneous. Bland Altman analysis was used to compare parameters. Comparing the values calculated by Breath-Sep with the data from the EVITA® using Bland-Altman analyses showed a mean bias of - 2.85% and 95% limits of agreement from - 25.76 to 20.06% for MV. For respiratory rate (RR) RR a bias of 0.84% with a SD of 1.21% and 95% limits of agreement from - 1.53 to 3.21% were found. In the cluster analysis of the 25th highest breaths of each group RR was higher using the EVITA®. In the mechanical subgroup the values for RR and MV the EVITA® showed higher values compared to Breath-Sep. We developed a computerized method for respiratory flow-curve based differentiation of breathing cycle components during mechanical ventilation with superimposed spontaneous breathing. Further studies in humans and optimizing of this technique is necessary to allow for real-time use at the bedside.