An Alberta-based tech company is reaching out to universities and machine learning specialists from around the world to contribute to its Caredemic project, an app that serves as an earlier screening process for the ongoing pandemic.
The app has received contributions from Western Economic Diversification Canada, Kinkaide Enterprises Inc., Universe Machine Corporation, and several anonymous donors. In order to get the project off the ground, Thomas Stachura, the chief executive officer of Pleasant Solutions, told IT World Canada in an email. In addition to Caredemic, Pleasant Solutions, a software developer, has multiple projects under its belt, including its Paranoid project, aimed at disabling smart homes’ prying ears.
“This is a company that has been around for a dozen years, and whose software NASA trusts to manage employee passwords. So does the U.S. Department of Energy. So does the staff at Buckingham Palace. We recently developed a line of digital privacy devices and launched those earlier this year. We have an extensive history in the tech industry, with most of our business in the B2B sphere,” Stachura wrote in the email.
With Caredemic, the screening process will measure different data points in a user, including his or her oxygen levels, pupil dilation, breathing rate, acoustics and volume, via the sensors that are built-in most smartphones. The oxygen level in the blood will be measured by a thumbprint scan, pupil dilation by the camera and breathing rate, acoustics and volume via the microphone.
This data will then be uploaded and fed into a machine learning model to generate a probability: is the user a likely candidate for self-isolation and/or definitive medical testing?
There’s no timeline for the app’s availability, but anyone will be able to download the Caredemic app on a smartphone themselves. Caredemic stresses the app is not intended to replace medical testing. The app will provide a valuable supplement to assist in early detection of COVID-19—even in cases with mild or asymptomatic infections—within reasonable degrees of true versus false positives. In symptomatic cases, the tool could generate better precision and more useful data than most medical personnel could achieve with an ordinary visual assessment.
The company says that the app works using its extensive data collection list based on the complexity of what the predicted early signs of disease. Therefore, different machine learning and data analysis teams around the globe are needed so that different algorithms for different insights can be tackled effectively.
“Based on the core data collected, every team will be able to create a “derivative” which is either a view of the data or metadata. Derivatives can chain to depend on each other, within the same Machine Learning team or across different teams. This is why we need at least 10 Machine Learning volunteer teams around the globe. We are actively seeking Machine Learning specialists to assist us. There is even prize money attached, as an extra incentive to participate,” Stachura wrote.
The company says when it comes to finding data correlations, the expertise needed for this is technical rather than medical.
“We need to have the confirmed positives identified by the algorithms. For example, the ‘acoustic signatures of pneumonia’ is more of a biotechnical question than a medical one. The website has a list of references that can be found here,” Stachura said.
The company plans to launch a global call for volunteers, both infected and healthy (16 years of age or older), to contribute baseline data for the initial machine learning phase of this project. In a completely anonymous process, participants will complete a brief survey, and then carry out Caredemic’s series of measurements. The company says it will need more than 10,000 volunteers to contribute to data in order for Caredemic to be effective. The questions and data collection points can be found here.
In order to ensure data privacy, the company says it will not collect volunteers’ names or GPS locations, phone’s IP addresses, serial numbers, or operating system IDs, and the information will be only used for the development of this app.
