Elimination of Drifts in Long-Duration Monitoring for Apnea-Hypopnea of Human Respiration

This paper reports a methodology to eliminate an uncertain baseline drift in respiratory monitoring using a thermal airflow sensor exposed in a high humidity environment. Human respiratory airflow usually contains a large amount of moisture (relative humidity, RH > 85%). Water vapors in breathing air condense gradually on the surface of the sensor so as to form a thin water film that leads to a significant sensor drift in long-duration respiratory monitoring. The water film is formed by a combination of condensation and evaporation, and therefore the behavior of the humidity drift is complicated. Fortunately, the exhale and inhale responses of the sensor exhibit distinguishing features that are different from the humidity drift. Using a wavelet analysis method, we removed the baseline drift of the sensor and successfully recovered the respiratory waveform. Finally, we extracted apnea-hypopnea events from the respiratory signals monitored in whole-night sleeps of patients and compared them with golden standard polysomnography (PSG) results.