Abstract
BACKGROUND: Cardiogenic oscillations in airflow can cause ventilator autotriggering during pressure support ventilation, potentially leading to inappropriate hyperventilation. A method to attenuate these oscillations in real time may help reduce autotriggering. MATERIALS AND METHODS: High-resolution airflow and ECG signals were collected from intubated surgical patients receiving pressure support ventilation. Singular spectrum analysis (SSA) was applied in a sliding-window format to generate a smoothed respiratory waveform. We quantified attenuation of cardiogenic oscillations using ECG-aligned timing, frequency-domain analysis, and reduction in cardiac-frequency spectral power. Waveform fidelity was assessed using respiratory-envelope correlation and root-mean-square error (RMSE). Computational feasibility was evaluated by measuring processing time per window. RESULTS: SSA substantially reduced cardiac-frequency spectral power (82-87% reduction) while preserving respiratory structure (correlation with respiratory envelope 0.92-0.94). Reconstruction error was modest (RMSE 0.08-0.11 normalized units). Computation time per 6-s window was 14-22 ms, supporting potential real-time use. Attenuation performance remained stable during changes in respiratory rate. CONCLUSIONS: Sliding-window SSA attenuated cardiogenic oscillations in patient airflow signals and preserved the dominant respiratory pattern. As a proof-of-concept, this approach shows potential for integration into autotrigger-suppression logic, though further validation in larger and more diverse populations is required.