MIT researchers have developed an algorithm that can diagnose coronavirus based on cough sounds
This development is based on an application thatcan identify respiratory disorders such as pneumonia and bronchitis by analyzing a person's exhalation. However, scientists warn that these algorithms are biased - they cannot detect diseases in most cases, but this development could help during the coronavirus pandemic.
Researchers trained a system based on dozensthousands of cough samples. Preliminary research has shown that sounds can be an indicator of how weak or strong a person's vocal cords are. They then trained a second model to discern emotions in a person's speech, and a third model was trained on a cough database to see changes in lung and airway function.
The Russians named the most polluted cities in the country
Researchers have collected over 6,000 records. Some of them, 2.5 thousand records, are associated with coronavirus, and 2.5 thousand records where patients coughed, but were not infected. The algorithm had to determine over and over again which group of people was infected with the coronavirus.
After combining the model, the team testedher on a thousand records from the set. The researchers claim that they were able to identify 98.5% of cases of coronavirus in humans by coughing alone. The scientists also claim that the algorithm correctly identified all asymptomatic coughs.
Researchers at Massachusetts Technologyinstitutes emphasize that this model should not be the only tool for diagnosing coronavirus. They hope to use it to develop a free pre-screening app based on their AI model. Scientists are already collaborating with several hospitals in the United States.
Read also
The annual mission in the Arctic has ended and the data are disappointing. What awaits humanity?
See the closest shots of the Sun's surface
Coronaviruses have learned to mimic human immune proteins and trick the body