This AI cough analysis tool is far less invasive than COVID-19 swab tests

And its accuracy rate is remarkably high, too.

A woman is seen wrapped in a blanket, coughing into her hand. She is holding a tissue and looks ill.

Imagine if you could accurately identify COVID-19 in an individual with the help of a smartphone app that only needs to listen to the person’s cough. Researchers at the University of Essex have released a study on a cough analysis tool, powered by artificial intelligence, that they designed to assess if someone has COVID-19. The team behind this application, DeepCough3D, says that the AI tool correctly spotted COVID-19 positive patients with 98.800.83 percent accuracy.

"Our proposed web tool and underpinning algorithm for the robust, fast, point-of-need identification of COVID-19 facilitates the rapid detection of the infection," the researchers say in their paper. "We believe that it has the potential to significantly hamper the COVID-19 pandemic across the world."

A less intrusive test — DeepCough3D's makers studied more than 8,000 audio samples of patients' coughing in April 2020. The group analyzed these samples in Mexico and Spain, and found that close to 2,000 of the patients they studied had tested positive for the coronavirus.

DeepCough3D categorizes coughs into three different classes based on their severity, accurately identifying the presence of COVID-19 at a 98 percent rate in a clinically-validated set of data. If this method of identifying COVID-19 in people goes mainstream, we might have a chance at undergoing tests that are far less physically invasive than the current swab tests. In the present method, medical experts conduct a nasopharyngeal swab analysis, which involves deeply probing your nasopharynx with a long and rather uncomfortable swab lodged through your nose.


Depending on whether you have a deviated septum and how sensitive your tissues are in your nose, you might find the swab test somewhat uneasy or you might be one of the people who have described this method as painfully sharp. With the Essex team's cough analysis, physical intrusion for testing is subtracted from the process entirely.

A helping hand for hospitals — The group says that its DeepCough3D tool is special compared to other cough analysis tools because of its remarkably high accuracy and sensitivity rates. The group’s cough severity classification is particularly beneficial to hospitals because it allows medical infrastructure to prioritize the allocation of ventilators based on a case's urgency.

According to the Essex scholars, the next plan for DeepCough3D is to carry out interventional or experimental studies, which means that the researchers will apply their AI tool's cough analysis to broader pools of patients. After that, the tool will need to be certified in order to be used officially.