Artefact Detection in Impedance Pneumography Signals: A Machine Learning Approach

Abstract

Impedance pneumography has been suggested as an ambulatory technique for the monitoring of respiratory diseases. However, its ambulatory nature makes the recordings more prone to noise sources. It is important that such noisy segments are identified and removed, since they could have a huge impact on the performance of data-driven decision support tools. In this study, we investigated the added value of machine learning algorithms to separate clean from noisy bio-impedance signals. We compared three approaches: a heuristic algorithm, a feature-based classification model (SVM) and a convolutional neural network (CNN). The dataset consists of 47 chronic obstructive pulmonary disease patients who performed an inspiratory threshold loading protocol. During this protocol, their respiration was recorded with a bio-impedance device and a spirometer, which served as a gold standard. Four annotators scored the signals for the presence of artefacts, based on the reference signal. We have shown that the accuracy of both machine learning approaches (SVM: 87.77 ± 2.64% and CNN: 87.20 ± 2.78%) is significantly higher, compared to the heuristic approach (84.69 ± 2.32%). Moreover, no significant differences could be observed between the two machine learning approaches. The feature-based and neural network model obtained a respective AUC of 92.77±2.95% and 92.51±1.74%. These findings show that a data-driven approach could be beneficial for the task of artefact detection in respiratory thoracic bio-impedance signals.

Data Augmentation and Transfer Learning for Data Quality Assessment in Respiratory Monitoring

“…The recording of the data followed the World Medical Association’s Declaration of Helsinki on Ethical Principles for Medical Research Involving Humans Subjects. More details about this dataset can be found in ( Blanco-Almazan et al, 2021 ; Moeyersons et al, 2021 ).…”

“…To obtain the class to which each signal belongs, four independent annotators were asked to assign a label to them. For this, the graphical user interface and the five classes defined in ( Moeyersons et al, 2021 ) were used. The classes 1 (Excellent signal quality), 2 (Good signal quality), 3 (Average signal quality) and 4 (Bad signal quality) refer to the BioZ signal with respect to the reference system.…”

“…Afterwards, it was resampled to the original sampling frequency (16 Hz), which results in a signal with a slower breathing rate. This transformation does not affect the labels of the data, given that in the guidelines for annotating the respiratory signals presented in ( Moeyersons et al, 2021 ), the critical time frame to consider a signal of average or bad quality is the 16.6% (i.e. 10 and 5 s for the COPD and BS signals, respectively) of its length.…”

“…In order to overcome the difficulties related with the removal of artefacts, some works have implemented signal quality indication approaches to identify the parts of the data that are useful for analysis ( Orphanidou et al, 2015 ; Johnson et al, 2015 ; Castro et al, 2016 ; Castro et al, 2018 ; Charlton et al, 2021 ; Moeyersons et al, 2021 ). In general, the quality indication for electrocardiogram and photoplethysmogram signals, for the estimation of heart rate, has been investigated further than the quality indication of other signals, such as respiration.…”

“…Previously, a quality index for respiratory signals was developed by Charlton et al (2021) , which then was compared to a machine learning framework for quality assessment by Moeyersons et al (2021) . Rozo et al (2021) presented the results of applying transfer learning to the previous framework.…”