Wi-Pulmo: Commodity WiFi Can Capture Your Pulmonary Function Without Mouth Clinging

• Published in: IEEE internet of things journal Vol. 12; no. 1; pp. 854 - 868
• Main Authors: Zhao, Peng, Huang, Jinyang, Zhang, Xiang, Liu, Zhi, Yan, Huan, Wang, Meng, Zhuang, Guohang, Guo, Yutong, Sun, Xiao, Li, Meng
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2025; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Accuracy; Algorithms; Contact; Correlation; Deep learning; Dynamics; Effectiveness; Exhalation; Lung; Lung cancer; Machine learning; Noise; Older adults; Pepper spray; Portable equipment; Pulmonary function; Pulmonary functions; Respiratory diseases; Spirometers; Spirometry; the elderly; Tumors; variational encoder-decoder (VED) network; WiFi sensing; Wireless fidelity
• ISSN: 2327-4662; 2327-4662
• DOI: 10.1109/JIOT.2024.3470321

Pulmonary function testing is a crucial examination for respiratory diseases. Current medical spirometers are bulky and inconvenient, while available portable spirometers are extremely expensive and often lack accuracy. Furthermore, both devices require direct contact, inevitably increasing the cross-infection risk. To tackle these challenges, we propose Wi-Pulmo, an end-to-end deep learning-based Wireless System that utilizes WiFi channel state information (CSI) to provide contact-free, convenient, cost-effective, and precise pulmonary function testing outside the clinical setting. Based on the analysis of thoracic and abdominal movement patterns, Wi-Pulmo first validates the feasibility of using WiFi to estimate pulmonary function. Then, Wi-Pulmo designs an efficient fine-grained sensing quality-based algorithm for complete exhalation segmentation. Additionally, a relevant interference-tolerant learning algorithm based on variational inference is proposed to accurately map the CSI of WiFi signals to pulmonary function. Extensive experiments achieved average monitoring error rates of 2.59% for normal subjects in daily scenarios and 5.87% for real patients in tertiary hospitals over a two-month period. These satisfactory results demonstrate the strong effectiveness and robustness of Wi-Pulmo. Furthermore, our findings in clinical reveal a close correlation between chronic diseases and pulmonary function.