Open Collections

Verma, Shreya

University of British Columbia

Measuring vital signs is crucial in healthcare as they provide essential information about a human body's physiological state. Particularly, the initial months following birth are crucial for the survival and well-being of newborns. Existing methods face challenges such as estimating incorrect respiratory rate values in low-light conditions, the inability to detect shallow breathing, overestimating respiratory values in darker skin tones and the need for contact sensors that may disturb sensitive infants. In this study, we propose a novel video-based respiratory rate estimation method for infants in the Neonatal Intensive Care Unit. Our method overcomes challenges by utilizing non-contact depth-sensing camera technology, which mitigates the impact of skin pigmentation on measurement accuracy. We conducted a study with 13 infants, recording uninterrupted 4-hour video sessions. We evaluated the algorithm's performance and robustness in various light conditions by analyzing randomly selected videos. We compared the respiratory rate estimated by our depth-sensing camera method with both a nurse practitioner's manually annotated reference respiratory rate and the respiratory rate estimated from red-green-blue (RGB) videos of the infants recorded during the same period as the depth videos. Our results record a bias of 0.125 bpm and a root mean squared difference accuracy of 1.058 bpm. Moreover, our proposed method exhibits a 26% increase in accuracy compared to the RGB camera-based method for respiratory rate estimation in infants. The Pearson correlation coefficient of 0.87 (p < 0.001) indicates a strong correlation, and the intraclass correlation coefficient of 0.77 (95%CI: 0.68 - 0.84 ) demonstrates good reliability. The Bland-Altman analysis of the proposed algorithm shows higher agreement between the values estimated by our developed method and the visually counted respiratory rate than the agreement between the respiratory rate obtained from the impedance sensors and the reference respiratory rate, the agreement between a former EVM-based method, and the respiratory rate estimated during the RGB-camera-based study. Our preliminary study shows that the proposed algorithm can successfully estimate the respiratory rate in the regular clinical care environment.

Text

2024-05-15

Attribution-NonCommercial-NoDerivatives 4.0 International

http://hdl.handle.net/2429/88291

Electrical and Computer Engineering

University of British Columbia

UBCV

http://creativecommons.org/licenses/by-nc-nd/4.0/