A group of Turkish scientists has designed an artificial intelligence (AI)-based system that can detect the coronavirus with 90% accuracy and is able to diagnose the infection within seconds, AzVision.az reports citing DailySabah.
The system is composed of a device scientists call the “electronic nose,” which is fed data via AI algorithms and taught to examine the breathing patterns of hundreds of patients.
Currently a prototype, scientists plan to introduce it at hospitals for faster detection of the coronavirus.
The project, conceived by professors Gökhan Silahtaroğlu and Mesut Yılmaz, actually started life as a way to determine honey quality and bee health. However, when the pandemic broke out last year, Yılmaz and Silahtaroğlu decided to improve it for use in COVID-19 diagnosis.
“We first started by putting sensors measuring air quality in beehives and feeding data from there to AI. The sensors were already used in cancer and tuberculosis detection and we decided to repurpose them to detect the coronavirus. Using sensors, we derived elements of breath and digitized them. By matching them with polymerase chain reaction (PCR) tests, we succeeded in helping the AI learn it,” Silahtaroğlu told Demirören News Agency (DHA) Thursday.
He says the device can be used in the diagnosis of other diseases as well, with data provided by patients.
The device was tested on 300 volunteers, who applied to hospitals for PCR tests and a comparison with the results demonstrated its success, the scientists say.
Mesut Yılmaz says apart from hospitals, it can be used at venues, like shopping malls, schools and other crowded places that require a fast diagnosis.
“People are given balloons and asked to inflate them. The air inside the balloon is then read by the device, which examines chemical activity in the lungs,” said Yılmaz explaining how it works.
Turkish scientists had mobilized for developing a vaccine against coronavirus while some researchers are working on drugs to eliminate the virus. A similar "breath test" is also being developed by scientists at Bandırma Onyedi Eylül University.