Pulmonary Screener: A Mobile Phone Screening Tool for Pulmonary and Respiratory Disease

Pulmonary and respiratory diseases comprise a large proportion of the global disease burden, responsible for both mortality and disability. This burden is especially concentrated in the developing world, where air pollution levels are generally high and resources for diagnosing these diseases are very limited. Health workers and many general practitioner doctors in developing countries are not trained to diagnose pulmonary diseases, leading to high rates of misdiagnosis and underdiagnosis. Motivated by this need, we have developed a mobile toolkit that can be used for screening and diagnostic guidance for three of the most common pulmonary and respiratory diseases (Asthma, Chronic Obstructive Pulmonary Disease (COPD) and Allergic Rhinitis (AR)). The toolkit consists of a mobile phone app, known as Pulmonary Screener, which is used in conjunction with a low-cost (peak flow meter. Machine vision software enables the phone camera to automatically track and capture the reading from the peak flow meter without the need for any electronics, Bluetooth radio, or batteries. Using the peak flow meter reading as well as an integrated clinical questionnaire, a machine learning model is then used to calculate the individual probabilities of a patient having a specific pulmonary disease (Asthma, COPD, AR, other) or co-morbidities (Asthma + AR, COPD + AR). The machine learning models used in the application were trained using diagnostic data from 325 patients collected at a pulmonary clinic over the past 3 years. Based on 50 iterations of a held-out test set, the Pulmonary Screener achieves accuracy (AUC) values above 0.90 for all diseases and combinations, with the exception of Asthma with AUC = 0.84. Sensitivity and specificity values were for all diseases was also greater than 0.90. To our knowledge, this is the first clinically validated mobile tool that is capable of diagnosing multiple pulmonary diseases in a single app. Future versions of the Pulmonary Screener will make use of ongoing data collection to expand support for infectious diseases as well, including pneumonia and tuberculosis, and include features extracted from auscultation and cough sounds.