AI Technology Can Detect COVID-19 Through Cough Sound
Artificial intelligence (AI) technology has now been implemented in various fields to facilitate work. The application of AI technology is also anticipated to be able to stop the spread of COVID-19 that currently rules the world.
A recent study from the Massachusetts Institute of Technology (MIT) revealed that COVID-19 can be detected using AI through cough sounds only. The technology, which was originally developed to detect symptoms of penumonia and asthma, can also detect COVID-19 in patients without symptoms (OTG).
Not only does it detect covid-19 symptoms in people who have symptoms and have no symptoms, similar algorithms are also developed to detect Alzheimer's disease. Brian Subirana, one of the AI researchers, said that researchers are investigating whether algorithms that detect symptoms through coughing can be useful in detecting Alzheimer's and COVID-19.
"There is evidence that infected patients experience some similar neurological symptoms, such as temporary neuromosular disorders," Brian said.
Researchers from MIT designed an AI that has three layers. First, the basic resNet50 algorithm to measure the strength of the vocal cords. Second, technology to determine emotional level, and the third is technology to detect anomalies in the respiratory system.
MIT researchers conducted AI research to detect COVID-19 through coughing sounds since April. In the process of making it, the researchers collected as many cough sound samples as possible, including from covid-19 sufferers who have symptoms and without symptoms.
To collect as many cough sound samples as possible, MIT researchers created a website specifically for participating participants to submit their cough sounds while filling out health surveys. Currently, researchers have collected as many as 70,000 cough sound samples with 2,500 of them coming from COVID-19 patients.
Using 2,500 covid-19 cough sound recordings, including OTG patients as well as 2,500 healthy people coughing sounds, the researchers took 4,000 samples to train AI algorithms.
Meanwhile, 1,000 residual samples from a combination of covid-19 and healthy patient cough sound data were used to see the accuracy of AI in detecting. MIT researchers found that this AI can detect COVID-19 with an accuracy rate of 98.5 percent, while detection in OTG has 100 percent accuracy.
The researchers also claim that the coughing sound of people exposed to COVID-19 is either symptomaanical or does not actually have a difference with a healthy person. They continue, the human ear can't hear it, but their designed AI technology can capture it.
With high accuracy, MIT researchers wanted to pin the technology to a smartphone app that could be downloaded for free. But before it can be widely used, it must be approved first by the United States Food and Drug Administration (FDA).
If successfully approved and released, then everyone can use the app to record their cough activity every day and detect independently if they are exposed to COVID-19 or not.
"The implementation of this application can reduce the spread of covid-19 virus, if people use it before going to school, factory, or restaurant," brian said.
