Humankind is facing an unprecedented disaster. I believe that we, as IT professionals, should act to help curb the pandemic. Below is a project idea. I invite anyone with the needed capabilities and resources to try it out.
First, we design a web platform, a website or an app, to allow people to update their symptoms continuously. Global Initiatives such as https://coronastatus.us and https://coronastatus.fr enable people to report their health condition related to the virus anonymously. Such websites are useful for the government to predict the corona spread, but for diagnosing, we probably need refined information. Any info that would help a physician is helpful, e.g., “day 1 – I got a sore throat; day 2 – I got a dry cough; day 3 – My body temperatures is 38 degrees.” The web platform should also allow people to input CT scans, blood tests, or coughing sounds, etc.
Then, we learn from the data via artificial intelligence, e.g., by building neural networks. The learned model allows us to make a diagnose. People will be motivated to use such a platform if they can be provided a free diagnosis that complement doctors’ decisions. I expect that the diagnosis can be highly accurate once a large amount of data has been collected. As an example, AI models already detect cardiovascular disease with a success rate of 98.5%, outperforming cardiologists.
Related to this project, Hospital Israelita Albert Einstein in Brazil applied AI to diagnose COVID-19 cases based on data from 235 patients. Most likely, their results can be improved with data collected from the web platform data, and in fact, hospital-provided and web data should merge for synergies. EPFL researchers have developed an AI-based corona diagnosing that can listen to coughing and indicate whether the patient has COVID-19. The app has an accuracy rate of 70%. Perhaps, the accuracy can be enhanced with the data we would ask users to submit to the web platform.