Abstract—Patient-Ventilator Asynchrony (PVA) is a common cause of ventilation-related medical complications and are traditionally only able to be reliably identified by trained clinicians. The need for constant monitoring and limited access to trained experts are major challenges in managing PVA, both of which can potentially be solved by automating the detection process. In this research, we propose a new data-driven approach to PVA detection using several similarity and randomness measures, including how unusual a time window is in the series and randomness of the time window. We found that all these similarity or randomness measures can be estimated with variants of the highly efficient Matrix Profile (MP) algorithm, and that one base routine can be repeated to generate all the features used in classification. We show that MP-based features, when used in combination with basic statistical and spectral features, can achieve an F-2 score of over 0.9 for two classes of PVA events in a sample of participants with moderate to high rate of PVA occurrence.
 

Index Terms—patient-ventilator asynchrony, matrix profile, anomaly detection

Cite: Chenyang Wang, Uwe Aickelin, Ling Luo, Goce Ristanoski, Mark E. Howard, and David Berlowitz, "Patient-Ventilator Asynchrony Detection via Similarity Search Methods," Journal of Advances in Information Technology, Vol. 13, No. 1, pp. 15-20, February 2022. 

Copyright © 2022 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.