A portable device for detecting excessive water in lungs

Abstract

Pulmonary edema (excessive water in lungs) can cause pulmonary hypertension, pleural effusion and acute case can be fatal. Currently excessive water in lungs is detected by medical devices using X-ray, CT scan or serum biomarker. Previous research works showed that acoustic method with machine learning algorithms can be applied. This project aimed to implement the previous machine learning algorithms using acoustic models on Android platforms. Because Android smart phones claim huge proportion of smartphone market and most of them are capable of signal real-time processing. During the project, Matlab scripts and a Perl script were written to prepare training data for Android application. Three-level Haar wavelet transform was used for feature extraction of lung sound recordings and kth-nearest-neighbor was used for classification in machine learning section. An Android Application was built for algorithm testing and real-time processing.