Research Article
Assessment of COVID-19 from Features Extraction of Exhaled Breath Using Signal Processing Methods
Exhaled breath waveforms, utilized in ventilation monitoring, are aimed at upgrading into COVID-19 disease screening. An algorithm for valid exhaled breath waveform segmentation and feature computation is developed to identify COVID-19 infection using exhaled breath patterns for distinguishing a COVID and non-COVID condition. Two minutes of exhaled breath patterns were recorded using a device and a nasal cannula sampling tube, resulting in the collection of exhaled breath waveforms from each subject. The developed algorithm is utilized to evaluate the valid exhaled breath waveforms and compute the features classified to distinguish COVID and non-COVID conditions. Slope e2, activity e2, and intersection angle of expiration and inspiration phases showed p-values of 0.000, denoting the strong significant difference between COVID and non-COVID conditions. The statistical analyses revealed p-values of 0.039, 0.008, and 0.024 for area e2, mobility of e2, and complexity e3, indicating their significance in differentiating the COVID-19 condition from the non-COVID condition. The slope, area, and intersection angle, as significant features, showed good predictive power for compliance with p-value analysis, with area under the receiver operating characteristic curves of 0.667, 0.693, and 0.775. The slope of e2, the area of e2, and the intersection angle of expiration and inspiration phases are identified as the promising features to be chosen in discriminating the COVID and non-COVID conditions.
see more