Predictive Modeling of Lung Water Content with Machine Learning and RF Sensor

This work addresses the challenges of early detection of pulmonary edema by predicting lung water content from scattering parameter data using two similar RF sensors. Printed microwave antenna, used as the sensor, collects data for the reflection coefficient as well as the transmission coefficient when placed on simulation blocks emulating the human thorax region over the lungs. Data for several demographic groups was substituted in the simulation blocks, and machine learning models were trained using the available data to build a dataset to predict the severity of pulmonary edema. This is believed to be the first of its kind approach in demonstrating the effectiveness of using microwave technique in early detection of pulmonary edema.