Machine Learning-Assisted Portable Electrical Impedance Tomography Enables Accurate Lung Function Assessment

Respiratory diseases have led to millions of deaths in the world (e.g., 3.5 million for chronic obstructive pulmonary disease (COPD)). While early detection can significantly improve patient outcome, choices for wide-spread, affordable screening tools are limited. Current clinical standard of lung function assessment relies on in-hospital spirometry, yet portable spirometers can only provide limited knowledge for the localization of respiratory pathological conditions which is crucial for personalized assessments and treatment. Meanwhile, electrical impedance tomography (EIT) has gained traction due to its affordability, safety, and portability. However, application in assessing lung function has traditionally been limited by heavy and bulk machinery. In this study, we present a novel system for evaluating lung function using a portable EIT system. Combined with machine learning, the system offers the ability to monitor lung function caused by conditions such as asthma, COPD, and other respiratory disorders. We also leveraged the power of machine learning and EIT to offer a clinically valuable alternative for affordable lung regional-based assessment. Our findings demonstrate good correlation with pulmonary functional test (R² = 0.619 (FEV1); 0.646 (FVC); 0.538 (FEV1/FVC), all p < 0.01) while providing interpretability into global and regional lung function prediction and potentially diagnosed COPD as in a clinical setting (sensitivity = 82%; specificity= 73%), highlighting the potential as an affordable, non-invasive tool for both global and regional respiratory assessment.Clinical RelevancePortable, affordable EIT system aided by machine learning, enables home and community-based monitoring and screening for multiple respiratory conditions and provide insights to regional pulmonary characteristics.