Paper Type
Complete
Abstract
Scalable respiratory monitoring is challenging in digital health systems when wearable models must operate reliably for new users without calibration. Although prior approaches achieve high accuracy under personalized conditions, they introduce scalability barriers that limit deployability. This proof-of-concept study adopts Design Science Research perspective to develop and evaluate a calibration-free tidal volume estimation artifact intended for plug-and-play integration in scalable digital health systems. The proposed solution combines inter-individual normalization, nonlinear modeling, and subject-independent validation using a leave-one-subject-out (LOSO) protocol to approximate real-world deployment conditions. Empirical results demonstrate that the nonlinear model significantly outperforms linear regression, achieving a pooled MAPE of 13.96% and Pearson correlation of 0.781 while maintaining stability across diverse breathing rates and motion scenarios. Beyond predictive performance, the primary contribution lies in reframing personalization removal as a system design objective. By demonstrating reliable cross-participant estimation without calibration, this work advances Health Information Systems through a scalable respiratory analytics component.
Paper Number
1684
Recommended Citation
Pawłowski, Szymon and Szymanski, Julian, "Calibration-Free Tidal Volume Estimation for Scalable Digital Health Systems" (2026). AMCIS 2026 Proceedings. 12.
https://aisel.aisnet.org/amcis2026/sig_health/sig_health/12
Calibration-Free Tidal Volume Estimation for Scalable Digital Health Systems
Scalable respiratory monitoring is challenging in digital health systems when wearable models must operate reliably for new users without calibration. Although prior approaches achieve high accuracy under personalized conditions, they introduce scalability barriers that limit deployability. This proof-of-concept study adopts Design Science Research perspective to develop and evaluate a calibration-free tidal volume estimation artifact intended for plug-and-play integration in scalable digital health systems. The proposed solution combines inter-individual normalization, nonlinear modeling, and subject-independent validation using a leave-one-subject-out (LOSO) protocol to approximate real-world deployment conditions. Empirical results demonstrate that the nonlinear model significantly outperforms linear regression, achieving a pooled MAPE of 13.96% and Pearson correlation of 0.781 while maintaining stability across diverse breathing rates and motion scenarios. Beyond predictive performance, the primary contribution lies in reframing personalization removal as a system design objective. By demonstrating reliable cross-participant estimation without calibration, this work advances Health Information Systems through a scalable respiratory analytics component.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
SIGHealth