Qurra : an Offline AI-based Mobile Doctor

Abstract

The recent global health pandemic has shifted the way healthcare is provided. Healthcare providers are overwhelmed with hospitals' beds being occupied and concerned about well being of visiting patients in need of regular checkups. Hence, innovative mobile-based healthcare solutions are needed. This work named Qurra or in Arabic presents a real-time solution that uses pre-trained built-in machine learning (ML)-models on mobile devices for convenient health checkups. The Qurra application operates by sampling data from various mobile sensors and uses the sampled data as input to different machine learning modules to produce a meaningful health diagnosis. The developed modules are the heart rate and cough detection. These modules are the focus of this paper. The approach used in this work relies on externally ML- based trained models that are ported to the application. Then, sensory inputs are tested against these pre-trained models, locally computed and analyzed on the mobile phone. The results are displayed to the user in real-time. This approach of having the models embedded to the mobile phone eliminates the need for internet connectivity. Moreover, the developed system is compared with a third party application in addition to its native model on a desktop computer. Also, the app has fast overall processing time of approximately 2-3.5 seconds.