[ad_1]
We won’t surprise you with the announcement that our smartphones, and the owners of the apps on them, know a lot about us. But when I happen to be fed up with Kovid-19 and my smartphone has learned it before I have, it already sounds like science fiction, Deutsche Welle reports.
Except it’s not science fiction, it’s just science. A new mobile app may soon appear on the market, diagnosing asymptomatic Kovid-19 disease only by the owner’s cough and speech.
Nearly 100 percent accurate
The application now recognizes infected people with great precision, and now its authors are looking for a way to reduce the percentage of misdiagnoses in healthy people. If successful, the new application can complement existing mobile applications to record contacts with virus carriers.
The idea for the new application belongs to three computer scientists from the Massachusetts Institute of Technology (MIT). In two months, Jordi Lugarta, Ferran Hueto and Brian Subriana analyzed more than 2,000 audio recordings of a total of 5,320 people. The records contained coughing and speaking. The researchers were looking for certain acoustic “biomarkers,” that is, characteristic sound marks. They already had experience with “biomarkers” in question from previous tests in Alzheimer’s patients.
The results so far are promising. “In people who have been officially diagnosed with coronavirus, the accuracy is 98.5%,” the authors wrote in a paper they proposed for publication in the IEEE Open Journal of Engineering in Medicine and Biology.
In other words: in proven sick people, the application achieves an accuracy of almost 100 percent. In asymptomatic patients who have not been previously evaluated for coronavirus infection, the application even recognizes 100% of those infected. The problem is that artificial intelligence is mistaken in the opposite direction: almost 20% of the people who, according to the application, are sick, are actually healthy.
Free, fast, always available
It is this deficit that must be addressed. Because otherwise, if the application goes into massive use, a large number of scared healthy people will be tested, as a result of which the capacity of the laboratory will quickly run out. Greater precision can be achieved by entering more information into computers, from which artificial intelligence can more accurately analyze the sound of different types of cough.
The authors are convinced that their application can become a free, always available, instantaneous and widespread tool for the diagnosis of Kovid-19 disease, which is otherwise asymptomatic.
[ad_2]
