Lung Disease Self-screening Using Deep Learning and Mobile Apps for Telehealth Monitoring

Based on data from the World Health Organization (WHO), lung disease is the third largest killer in the world. Therefore, this research aims to design a lung disease screening (LUDESC) system based on deep learning to monitor lung disorders early on using mobile applications for telehealth care. The proposed system comprises a digital stethoscope to hear lung sounds that can be directly connected to a smartphone with an audio jack connector. At the same time, the DL model embedded in the mobile apps allows users to recognize the lung condition of healthy, chronic obstructive pulmonary disease (COPD), bronchiolitis, bronchiectasis, or upper respiratory tract infection (URTI). The proposed application also provides appropriate health recommendations if the lungs experience abnormalities. The prediction model was built using ResNet50 1D trained on datasets downloaded from the International Conference on Biomedical and Health Informatics (ICBHI) 2017. This dataset consists of 5.5 h of total voice recordings from 120 patients and 6898 respiratory circulations. The simulation result shows that the classifier model based on the ResNet50 1D achieved an accuracy rate of 98%, precision of 91.3%, recall of 93.83%, and F1-score of 91%. The LUDESC testing results show very satisfactory values, indicating its ability to perform properly.