The most important part of the COVID-19 fight is to have its symptoms detected and quickly get isolated to stop the spread of it to other people. Asymptomatic people who have COVID-19 shows no symbol of the virus in them and they do not get tested as it stays unknown even to the one suffering from it and spread the infection to others unknowingly.
Researchers from MIT have found out that the sound of the cough differs in the asymptomatic people than the healthy ones. The difference in the sound of the cough is not detected by human ears but it is detected by Artificial Intelligence(AI). The research has been published in the IEEE Journal of Engineering in Medicine and Biology.
The AI model was schooled with thousands of human talks and cough samples. Later when the AI had tested various cough recordings, 98.5 percent of the tested coughs were having Covid-19 along with 100 percent from asymptomatic people who did not have any symptoms but tested positive for COVID-19.
The research team is currently trying to implement an app which could even help asymptomatic people to get their regular test for COVID-19. With the help of the app, the user can easily cough in their phone regularly and get a result of whether they are infected or not instantly. This would even save a lot of time for everyone.
Brian Subirana said: “The effective implementation of this group diagnostic tool could diminish the spread of the pandemic if everyone uses it before going to a classroom, a factory, or a restaurant.”
Before the pandemic, researchers were engaged in making up an AI that can inspect cough recordings to detect Alzheimer’s. The research involved a neural network called ResNet50. The training of the network was done in three major models –
- Thousand hours of speech by a human.
- A database of words applied by a human in various emotional states.
- Recording of coughs to identify changes in lung and respiratory performance.
All these data were mixed and a coat of noise was used to distinguish the harder coughs from weaker ones. The researchers fed recordings into the new AI and concluded that the new model can detect Alzheimer’s samples much better than the previous ones.
As the Coronavirus pandemic came into our life Subirana thought whether their AI can do the same task for COVID-19 patients too as there was some evidence developing that infected patients experience some neurological symptoms such as temporary neuromuscular impairment. The researchers started collecting cough recordings including the COVID-19 patients too.
The researchers gathered 70,000 recordings each having various coughs including 200,000 forced-cough recoding samples. According to subirana “It is the largest research cough dataset.” Nearly 2500audio recordings were submitted by people who had confirmation of COVID-19 along with the ones who were asymptomatic.
The recordings were then used to discriminate between Covid patients and healthy individuals. Researchers wrote in their published paper that their hard words showed “a striking similarity between Alzheimer’s and Covid discrimination.”
Source – mit