Home / Blog / AI tested me for Covid,… and I passed

AI tested me for Covid,… and I passed

This is my test result, based on voice analyses performed using AI algorithms. It looks like I have very low probability of beinig positive to Covid-19. Image credit: Carnegie Mellon Universityyou enter

Researchers at Carnegie Mellon University are working on an AI based test for Covid-19 that uses your voice pattern for analyses by comparing it to the voices of hundreds of thousands of other people, infected and not. You can participate in the trial and get yourself checked here.

Once you enter the testing stage you will see a disclaimer stating that this is an ongoing research and its results cannot be considered nor used as definitive answers.

What we know about AI is that its effectiveness largely depends on the latitude, number, of data sets upon which it can derive its conclusion. Clearly the more samples are available the more accurate the results.

The quickest way to learn about how this research works is to experiment by yourself, and by doing so you are also contributing to make results more accurate. For those that have no time to dedicate (it takes about 10 minutes to go through the on line test) here is how it works:

  1. you enter some basic data about yourself (age, weight…)
  2. you notify the presence of some basic symptoms (fever, cough, pain…)
  3. you are asked to cough three times in the mike
  4. you are asked to utter a few specific vowels sounds
  5. you recite the sequence of numbers 1 to 20
  6. you recite the alphabet

That’s it. You click “submit” and the systems will come back to you with a bar (see picture) indicating the probability you are positive (in my case it was low).

As I stated this is work in progress. What is really important is that scientists are looking at ways that can provide a quick answer at very low cost (basically zero) and that can be used as a first step in the assessment.

I trust that given a sufficient number of samples the AI can indeed provide a significant, meaningful result. It does not need to be 100% accurate, it should be enough to activate further action if needed.

How does it do this magic? Well, I guess that by scrutinising your voice utterance variations it can detect the existence of pockets of infection in your lungs. Those pockets are congested and contain less air. This shows as you expel the air form the lungs to keep uttering the various vowels. In normal pneumonia you have just one foci of infection and even if it is bilateral (affecting both lungs) it is localised (and spreads around). This is not so with Covid-19 that creates multiple foci of infection in both lungs. This creates a different pattern that through experience can be detected. The utterance of the numbers and alphabet creates the baseline against which the long continuous utterance of vowels is compared.

Now, I can imagine that this approach to Covid-19 test can evolve to the point of being embedded in our smartphone so that the testing is actually performed every time we take/make a call. That would increase enormously the amount of data and the sensitivity at personal level since it will be able to detect deviation in our voice patterns, hence prompting for a more accurate guided testing. Our own personal digital twin may take charge of this continuous testing by looking for specific patterns and notifying an AI diagnostic service when something suspicious is detected.

Difficult times are stimulating people all around the world and result in amazing answers that I am confident will help us through the current difficulties and most important will prepare us to stand better in front of those that will come in the coming years.

About Roberto Saracco

Roberto Saracco fell in love with technology and its implications long time ago. His background is in math and computer science. Until April 2017 he led the EIT Digital Italian Node and then was head of the Industrial Doctoral School of EIT Digital up to September 2018. Previously, up to December 2011 he was the Director of the Telecom Italia Future Centre in Venice, looking at the interplay of technology evolution, economics and society. At the turn of the century he led a World Bank-Infodev project to stimulate entrepreneurship in Latin America. He is a senior member of IEEE where he leads the Industry Advisory Board within the Future Directions Committee and co-chairs the Digital Reality Initiative. He teaches a Master course on Technology Forecasting and Market impact at the University of Trento. He has published over 100 papers in journals and magazines and 14 books.