AL assisted medical diagnosis of lung diseases using accurate symptoms visualization

Abstract

Mobile health tracking applications are one of the needed technologies that have a great potential to collect patients' health history effectively. Contact tracing, symptom tracking and assessment, and health information distribution are among the unmet needs of the healthcare and public health systems, which these apps can address. In this paper, we propose an AIassisted medical diagnosis multiplatform mobile application for lung diseases. The proposed application allows users to track their journey of symptoms as they develop, intensify, and resolve. With a few clicks, the application enables the user to log some symptoms, including headache, cough, and other symptoms and logs other health data like blood pressure and glucose levels automatically from device sensors and wearables through Google Fit integration. Recorded cough sounds are then classified based on spectrogram features using our Convolutional Neural Network on-device cough classifier with an 89% accuracy. Manually input and automatically detected symptoms are added to a timeline that visualizes the user's symptom history and intensity over time. Users can then navigate the symptom journey themselves or share it with a doctor when needed reducing the need for frequently asked questions by doctors and making it easier to identify and track their symptoms over a long period. Our results show that our proposed application can be used as a diagnostic tool to record symptoms anytime in a few clicks and generate history timelines to track the user's health. Our user-friendly application is an effective and organized way to document the user's health status regularly and can be shared with a doctor if necessary. © 2022 IEEE

Keywords Artificial Intelligence ,Cough classification, Mobile computing, Respiratory illness Diagnosis, Visualization