Automatic breath phase detection using only tracheal sounds

While automatic distinction between the two breath phases (inspiration/expiration) can be done easily using lung sounds' intensity, it is challenging to do the same using only tracheal breath sounds. The current acoustic flow estimation methods use tracheal breath sounds to estimate the amount of flow and the onset of breath but also use lung sounds for respiratory phase identification. It would be advantageous to have an automatic and accurate method to identify breath phases from the tracheal signal. One may argue that given the alternation of respiratory phases, breath phase identification from tracheal sounds would be an easy task if one knows the first phase. However, during breathing, an event such as apnea, swallowing, or coughing may change the alternating nature of breath phases. In this study we have investigated several parameters derived from the phase duration, the shape of the sound envelope within each phase, and the sound's intensity in each phase, to develop a reliable method to differentiate between the two respiratory phases using only tracheal breath sounds. We used data from 6 healthy individuals, without any history of pulmonary diseases at 4 different flow levels (shallow, tidal, medium and very high). The most prominent features were found to be those derived from the duration, area and shape of the sound envelope in each phase. With a voting equation using the three most prominent features, our proposed method has shown an accuracy of 93.1% with sensitivity of 93.4% and specificity of 92.8% for breath phase identification without the need for assuming breath phase alteration.