Joseph Prinable1,2, Peter Jones3, David Boland3, Alistair McEwan1, Cindy Thamrin2. 1. The School of Biomedical Engineering, University of Sydney, Darlington 2006, Australia. 2. The Woolcock Institute of Medical Research, University of Sydney, Glebe 2037, Australia. 3. The School of Electrical and Information Engineering, University of Sydney, Darlington 2006, Australia.
Abstract
The ability to continuously monitor breathing metrics may have indications for general health as well as respiratory conditions such as asthma. However, few studies have focused on breathing due to a lack of available wearable technologies. To examine the performance of two machine learning algorithms in extracting breathing metrics from a finger-based pulse oximeter, which is amenable to long-term monitoring. METHODS: Pulse oximetry data were collected from 11 healthy and 11 with asthma subjects who breathed at a range of controlled respiratory rates. U-shaped network (U-Net) and Long Short-Term Memory (LSTM) algorithms were applied to the data, and results compared against breathing metrics derived from respiratory inductance plethysmography measured simultaneously as a reference. RESULTS: The LSTM vs. U-Net model provided breathing metrics which were strongly correlated with those from the reference signal (all p < 0.001, except for inspiratory: expiratory ratio). The following absolute mean bias (95% confidence interval) values were observed (in seconds): inspiration time 0.01(-2.31, 2.34) vs. -0.02(-2.19, 2.16), expiration time -0.19(-2.35, 1.98) vs. -0.24(-2.36, 1.89), and inter-breath intervals -0.19(-2.73, 2.35) vs. -0.25(2.76, 2.26). The inspiratory:expiratory ratios were -0.14(-1.43, 1.16) vs. -0.14(-1.42, 1.13). Respiratory rate (breaths per minute) values were 0.22(-2.51, 2.96) vs. 0.29(-2.54, 3.11). While percentage bias was low, the 95% limits of agreement was high (~35% for respiratory rate). CONCLUSION: Both machine learning models show strong correlation and good comparability with reference, with low bias though wide variability for deriving breathing metrics in asthma and health cohorts. Future efforts should focus on improvement of performance of these models, e.g., by increasing the size of the training dataset at the lower breathing rates.
The ability to continuously monitor breathing metrics may have indications for general health as well as respiratory conditions such as asthma. However, few studies have focused on breathing due to a lack of available wearable technologies. To examine the performance of two machine learning algorithms in extracting breathing metrics from a finger-based pulse oximeter, which is amenable to long-term monitoring. METHODS: Pulse oximetry data were collected from 11 healthy and 11 with asthma subjects who breathed at a range of controlled respiratory rates. U-shaped network (U-Net) and Long Short-Term Memory (LSTM) algorithms were applied to the data, and results compared against breathing metrics derived from respiratory inductance plethysmography measured simultaneously as a reference. RESULTS: The LSTM vs. U-Net model provided breathing metrics which were strongly correlated with those from the reference signal (all p < 0.001, except for inspiratory: expiratory ratio). The following absolute mean bias (95% confidence interval) values were observed (in seconds): inspiration time 0.01(-2.31, 2.34) vs. -0.02(-2.19, 2.16), expiration time -0.19(-2.35, 1.98) vs. -0.24(-2.36, 1.89), and inter-breath intervals -0.19(-2.73, 2.35) vs. -0.25(2.76, 2.26). The inspiratory:expiratory ratios were -0.14(-1.43, 1.16) vs. -0.14(-1.42, 1.13). Respiratory rate (breaths per minute) values were 0.22(-2.51, 2.96) vs. 0.29(-2.54, 3.11). While percentage bias was low, the 95% limits of agreement was high (~35% for respiratory rate). CONCLUSION: Both machine learning models show strong correlation and good comparability with reference, with low bias though wide variability for deriving breathing metrics in asthma and health cohorts. Future efforts should focus on improvement of performance of these models, e.g., by increasing the size of the training dataset at the lower breathing rates.
Authors: Peter H Charlton; Timothy Bonnici; Lionel Tarassenko; David A Clifton; Richard Beale; Peter J Watkinson Journal: Physiol Meas Date: 2016-03-30 Impact factor: 2.833