A group of researchers from Carneige Mellon University, in Pennsylvania, USA, developed an application that claimed to be able to detect whether someone was infected with Covid-19. The detection is produced from cough and sound analysis using an algorithm.
Benjamin Striner, a master’s student who participated in this research said the software algorithm developed by his team will be very helpful to track the spread of the corona virus , although it still needs to be further investigated.
Striner and the team continue to collect voice data to improve the accuracy of the application that is currently named Covid Voice Detector. If it has been released, this application can be downloaded and installed on a smartphone or laptop.
Subtle signs through coughing
This application will serve as an indicator to measure the likelihood of someone infected with Covid-19. To detect it, the application will ask “patient” to cough several times, then say the vowel sound and read the alphabet.
After completing all the tests, the application will display a score in the form of indicator lines that will indicate whether someone is infected with Covid-19 or not based on algorithm calculations.
According to computer science professor Carneige Mellon, Rita Sigh who was involved in this project, the sound of coughing by Covid-19 sufferers was different from non-sufferers.
“(The virus) infects the lungs so badly that it affects breathing patterns and other vital parameters, and sufferers have strong voice signs,” explained Sigh.
Sigh is known to have for years developed an algorithm that can identify micro-signatures or subtle signs in the human body which he believes can reveal a person’s psychological, physiological, and medical data.
Application development is constrained because the researchers do it from their homes. Because, like most campuses in the world, the University of Carnegige Mellon was also closed due to the Covid-19 pandemic.
From their respective homes, they collected positive patient voice data for Covid-19 to practice the algorithm . They were also assisted by researchers from other countries to collect data.
The team not only collected sounds from positive patients from Covid-19 but also other virus sufferers.That is so that the algorithm is trained to detect sounds caused by different viruses. They also looked for videos of positive Covid-19 patient interviews from news sites.
“We have samples of healthy people, we have samples of people who might have contracted the flu, and also all the different cough records, how coughing out there can then lead to a differentiating point,” Striner explained.
Striner said, this application will give false positive results rather than false negative . False positives will show results that may be positive but that does not mean that someone is really infected with Covid-19, so they have to do laboratory tests.
Meanwhile, false negative might show negative results, but that does not mean the person is really not infected. “If you give false negative information , then they will walk around and can make a lot of people sick, rather than giving more false positive information , maybe some people get test results that they don’t need,” explained Striner.
Accuracy has not been tested
For now, it’s not certain how accurately this application detects Covid-19 infection. Both Striner and Sigh said that any test results from the application could not be used as medical advice. “The accuracy cannot be tested at this time because we have not conducted the verified verified trials that are needed,” Sigh said.
The application is also still under development and has not yet received approval from the US Food and Drug Administration (FDA) and the Centers for Disease Control and Prevention (CDC).
The main purpose of this application is to collect more voice data from both infected and healthy people to make the algorithm more intelligent at detecting.
So, for example there are application users who have a history of certain respiratory diseases, the algorithm can record how the sound of this type of disease.
“If later this application is used for public services, the results must still be verified by a professional medical team and through trials by related institutions such as CDC,” said Bhiksha Raj, professor Carneige Mellon who is also a member of the project research team.
He said, the application could help people to determine medical measures based on the results of the indicators. That means, this application will still not be as accurate as the results of the laboratory with more sophisticated and appropriate medical devices.
Even though it sounds interesting, Ashwin Vasan, a Columbia University Medical professor, is actually giving criticism. He objected if the application was actually released in the middle of a pandemic.
“Apart from the good intentions of the research team who helped the crisis, this news is not what we want to convey out there,” he said, summarized from Futurism, Wednesday (04/08/2020). He said, currently the medical teams in various countries lacked good test kits to diagnose the corona virus accurately.
In addition, medical staff also lack personal protective equipment (PPE) and ventilators for critical patients. “Let’s stay focused on that, especially as leaders in Washington don’t seem to be able to meet that most basic need,” he explained.