MIT researchers working on lightning-fast method to detect coronavirus

 

• Coronavirus, detected only by the sound of coughing. The revolutionary program that MIT researchers are working on

Researchers at the Massachusetts Institute of Technology (MIT) are working on a lightning-fast method to detect coronavirus - a test based solely on the sound of coughing.

MIT has released preliminary results of a test system that uses Artificial Intelligence (AI) to listen to the sound of coughing. With proper training, an AI algorithm can distinguish a "normal" cough from one infected with SARS-CoV-2, a differentiation that the human ear cannot make, The Economist reports in the podcast "Babbage" .

The program will use acoustic biomarkers, which measure vocal cord strength, lung performance and muscle degradation. A biomarker is a biological characteristic, which can be molecular, anatomical, physiological or biochemical. These characteristics can be measured and evaluated objectively, acting as indicators of a normal or pathological biological process.

To train the algorithm, MIT researchers used 4 different databases. The team also collected forced cough recordings through a website between April and May, developing what researchers claim to be the largest COVID-19 audio data set to date, with 70,000 recordings, of which 2,680 were sent by people confirmed to have COVID-19.

Initially, the MIT team developed AI models for this project from scratch, but reached a ceiling of approximately 70% accuracy. In one test, researchers trained their existing AI model of Alzheimer's disease with COVID-19 cough data and it worked. The model had an accuracy of 98.5% in detecting people found positive. In detecting asymptomatic people, the accuracy increased to 100%.

If, for example, such a program is installed in a restaurant where access will be based on a negative test, it can detect negative cases with 100% accuracy.

---------------------------------------

COURTESY universul.net

--------------