Virufy: donate your cough and help detect Covid-19
February 22nd, 2021
At XOOR we are collaborating with Virufy in the development of a web app to gather information and eventually be able to diagnose COVID-19 using cough sounds. The app works with artificial intelligence and for this stage of the project we are relying on people’s good will so they donate their coughs to improve disease detection effectiveness.
Once the expectations for detection precision are met, the next stage of the project could be launched, which contemplates disease diagnosis. The app will be available to hospitals, clinics and even villages without Internet access, to detect the presence of Covid-19 and so doctors can suggest the appropriate sanitary measures to heal the person and avoid spreading of the virus.
- [Visit Virufy’s website](https://virufy.org/app/welcome](https://virufy.org/app/welcome)
- Select language, the country from which you’re donating and press “next”
- Read information about Virufy and click “next”
- Read “consent for testing”, check the boxes and press “next”
- Read the health recommendations before recording and press “begin”
- Read the instructions to record correctly, press on the Mic button to begin coughing, pause when you are done and click “continue”
- Check that your cough was recorded correctly and press “done”
- Answer: Which of the following Covid-19 tests have you had done?
- Tell us When did you have your Covid-19 test done?
- State age and gender identity
- Report whether you smoked frequently at any point during the past 6 months?
- Select whether you are experiencing current Covid-19 symptoms
- Specify if you have any of the respiratory conditions listed or add a new one.
- Select any of the medical conditions listed you may have or add a new one. Check that you’re not a robot and press the “send form” button and YOU’RE DONE!
It’s a non profit organization made up of an interdisciplinary team of volunteers that facing the Covid-19 threat and amidst a global pandemic found a way to diagnose the disease using artificial intelligence.
The Virufy organization is made up of over 250 professionals and students from 22 universities in the U.S., Brazil, London, Spain, Japan and Canada, and supported by the Stanford Covid-Response Innovation Lab and One Young World. Here at XOOR we were called to collaborate in the development of the web app to bring this idea to life and to make it available across Latin America.
The organization, located in Silicon Valley, based themselves on some investigations to shape the project. They used a 2015 paper that analyzed the sound of coughs to classify pneumonia and asthma in children; a 2016 study that developed an algorithm for automatic diagnosis of whooping cough, and a 2008 test to identify respiratory infections in pigs.
In that sense, Cambridge University demonstrated that a binary learning classifier can detect patients positive for Covid-19 using breath and cough sounds with great accuracy.
Also, Carnegie Mellon University (CMU), Massachusetts Institute of Technology (MIT) mapped 18 voice characteristics that identify COVID-19 positive patients and trained a model to diagnose COVID-19 with a precision of 89.1%.
Finally, Virufy translated that data to an artificial intelligence algorithm that, when trained, can discern between a healthy cough and a Covid-19 cough.
Health professionals use breath and cough sounds to evaluate the health of patients and diagnose illnesses. In general, when we visit their office, doctors auscultate us and asks us to force a cough to listen to its characteristics.
Using that as a basis, Virufy developed a system that records those sounds and by using trained algorithms, it detects patterns with high sensibility and specificity for the classification of respiratory diseases.
As is the case with asthma and pneumonia, COVID-19 creates a unique respiratory signal in the throat and lungs. By feeding the algorithm with thousands and thousands of healthy and Covid-19 positive coughs, artificial intelligence learns to detect those signs and patterns and it can, therefore, discriminate and classify them.
According to the samples gathered currently in the U.S., Virufy can recognize within minutes with a precision greater than 90% a Covid-19 infection comparing one person’s cough to all other coughing records already in the database.
But in the rest of the world it has a precision of 80%. In order to use it effectively to diagnose Covid-19, the organization intends to achieve a precision of 90% or higher.
To reach that goal, Virufy expanded their borders thanks to the work of young professional volunteers across the globe. Its mission is to gather data from different regions in order to obtain a model for free Covid-19 diagnosis with better racial and geographical inclusion. To do so, it needs to broaden its database with the greater ethnic range and with the highest phonological differences possible for a better training of the artificial intelligence.
That's why, here at XOOR, we contribute to this cough donation campaign in Latin America, inviting everyone, whether healthy or with a respiratory illness or even with Covid-19 to open the app and leave their free sample in just 2 minutes.
One of the main concerns for doctors and infectious disease specialists across the world is the effectiveness and speed of Covid-19 detection to avoid spreading the virus -given its rapid rate of contagiousness-; and to avoid the collapse of the health system, and consequently, to reduce death rates.
Currently, present strategies to detect COVID-19 require testing people using expensive kits that are not always easy to access, specially in poor communities without free public health systems.
Once effectiveness parameters have been reached, Virufy will be useful to detect the disease for free across the world. You will only need a smart device such as a smartphone or tablet where you have previously downloaded the app with its corresponding database and you’ll be able to use it even in the farthest corners of the earth without needing an Internet connection.
Also, those coughs used to make the diagnosis will serve as samples that will feed the app’s database and, as such, they will optimize effectiveness in those regions.
Once the Covid-19 pandemic is over, the app could be used to diagnose several current contagious respiratory diseases, and future virus mutations that could affect the world using the same principle.
Tests could be designed for communities that fear the propagation of a specific virus in their region. And it could even be used to detect influenza or the common cold.
If you have any questions or would want to collaborate with us in any way, you can reach us at firstname.lastname@example.org and we will contact you shortly.