BreathFinder: A Method for Non-Invasive Isolation of Respiratory Cycles Utilizing the Thoracic Respiratory Inductance Plethysmography Signal

• Main Authors: Holm B, Borsky M, Arnardottir ES, Serwatko M, Mallett J, Islind AS, Óskarsdóttir M
• Format: Article in Journal/Newspaper
• Language: English
• Published: Dove Medical Press 2024
• Subjects: respiratory analysis; breath detection algorithm; sleep analysis; breath segmentation; respiratory cycle isolation; Psychiatry; RC435-571; Neurophysiology and neuropsychology; QP351-495; Iceland
• Online Access: https://doaj.org/article/1643fb3280c5493788309166548476ac

Benedikt Holm,1 Michal Borsky,1 Erna S Arnardottir,2,3 Marta Serwatko,2 Jacky Mallett,1 Anna Sigridur Islind,1 María Óskarsdóttir1 1Reykjavik University, School of Technology, Department of Computer Science, Reykjavik, Iceland; 2Reykjavik University, School of Technology, Sleep Institute, Reykjavik, Iceland; 3Landspitali, The National University Hospital of Iceland, Reykjavik, IcelandCorrespondence: Benedikt Holm, Email [e-mail]: The field of automatic respiratory analysis focuses mainly on breath detection on signals such as audio recordings, or nasal flow measurement, which suffer from issues with background noise and other disturbances. Here we introduce a novel algorithm designed to isolate individual respiratory cycles on a thoracic respiratory inductance plethysmography signal using the non-invasive signal of the respiratory inductance plethysmography belts.Purpose: The algorithm locates breaths using signal processing and statistical methods on the thoracic respiratory inductance plethysmography belt and enables the analysis of sleep data on an individual breath level.Patients and Methods: The algorithm was evaluated against a cohort of 31 participants, both healthy and diagnosed with obstructive sleep apnea. The dataset consisted of 13 female and 18 male participants between the ages of 20 and 69. The algorithm was evaluated on 7.3 hours of hand-annotated data from the cohort, or 8782 individual breaths in total. The algorithm was specifically evaluated on a dataset containing many sleep-disordered breathing events to confirm that it did not suffer in terms of accuracy when detecting breaths in the presence of sleep-disordered breathing. The algorithm was also evaluated across many participants, and we found that its accuracy was consistent across people. Source code for the algorithm was made public via an open-source Python library.Results: The proposed algorithm achieved an estimated 94% accuracy when detecting breaths in respiratory signals while producing false positives that ...