Artificial Intelligence to Detect Coronavirus Infection Among Individuals Without Actual Test – The Weather Channel

Original article was published on artificial intelligence

A doctor collects a throat swab specimen for the test of the novel coronavirus that causes COVID-19, at Kurla in Mumbai.

(Deepak Turbhekar/BCCL, Mumbai)

As the novel coronavirus pandemic COVID-19 continues to spread across the globe, researchers are racing against time to find possible preventive measures, tests and cures to arrest the spread. While the pandemic enters the stage of community spread in many parts of the world, countries are running short of essential medical kits to test sufficient numbers of people.

Testing is the need of the hour, and to catalyse the pace of testing, scientists have now developed an artificial intelligence-based diagnostic tool. The incredible new tool can help predict if an individual is likely to have COVID-19 disease, based on the symptoms they display. The discovery was recently published in the journal Nature Medicine.

Researchers developed the artificial intelligence-based model using data from an app called ‘COVID Symptom Study’. So far, the app is said to be downloaded by about 33 lakh people globally. The users report their health status daily on the apps, and according to the paper, the app collects data from both asymptomatic and symptomatic individuals. Besides, it tracks in real-time the disease progression by recording self-reported health information daily.

To develop the AI-based prediction system, researchers examined the data collected from about 25 lakh people in the United Kingdom and the United States between March 24 and April 21. These users actively used the app regularly to add their health status.

Based on the user data on symptoms and health status of users, the AI-based models predict who might have COVID-19. The model also uses the actual test results of the people who have been tested positive. The tool also looked into information such as test outcomes, demographics, and pre-existing medical conditions.

The research team analysed several symptoms of COVID-19, which are most likely to give positive results. These key symptoms include cold, flu, fever, cough, fatigue. Moreover, they also found loss of taste and smell, as a common characteristic of COVID-19 disease.

When the AI-based model was applied to over 800,000 app users who displayed exact symptoms—revealed about 17.42% of these people were likely to have coronavirus. Also, the tool has been proven beneficial in recognising patients who have developed mild symptoms. This could help stop the spread of the virus by making the people aware that they might be potential carriers.

The most valuable feature of this AI model is that it can predict the COVID-19 symptoms without patients getting the actual test. Particularly at the time of a pandemic—the app could prove to be of significant value for highly populated countries like India.