Quality Assurance for Home Spirometry using Machine Learning

Tools

Gardiner, Darcey, Rogers, Harry, Lines, Jason, Wilson, Andrew and Aung, Min Hane (2025) Quality Assurance for Home Spirometry using Machine Learning. In: IEEE Symposium on Computers and Communications (ISCC). The Institute of Electrical and Electronics Engineers (IEEE). (In Press)

[thumbnail of ICTS4eHealthPaper2025]
Preview
 PDF (ICTS4eHealthPaper2025) - Accepted Version
Download (424kB) | Preview

Abstract

Spirometry is used for evaluating lung function, playing a crucial role in assessing lung health and monitoring treatment effectiveness. Numerous studies demonstrate the potential of Machine Learning algorithms to match human experts in spirometry classification, although most approaches depend on custom data pre-processing and complex model architectures. Therefore, we apply efficient Time Series (TS) classifiers to quickly and computationally assure spirometry signal quality, enabling real-time deployment. Seven classifiers were implemented to classify spirometry curves as ‘Acceptable’ or ‘Not Acceptable’, with performance referenced against results from similar studies. The best-performing classifier was FreshPRINCE, a TS method combining TSFresh feature extraction with Rotation Forest classifier. The FreshPRINCE model achieved an accuracy of 0.9449, precision, recall and F1 Score of 0.9745, 0.9586 and 0.9665, consistently matching and sometimes outperforming more complex models. These findings suggest models, such as FreshPRINCE, could streamline spirometry analysis, reducing computational burden, whilst maintaining classification performance.

Item Type: Book Section
Faculty \ School: Faculty of Science > School of Computing Sciences
Faculty of Medicine and Health Sciences > Norwich Medical School
UEA Research Groups: Faculty of Medicine and Health Sciences > Research Groups > Respiratory and Airways Group
Depositing User: LivePure Connector
Date Deposited: 26 Aug 2025 10:30
Last Modified: 26 Aug 2025 10:30
URI: https://ueaeprints.uea.ac.uk/id/eprint/100238
DOI:

Downloads

Downloads per month over past year

Actions (login required)

View Item View Item