When an epidemic spreads, it’s vital to alert people who may have been infected but are yet to show symptoms. This process is called contact tracing, and across the world officials are spending long nights and weekends interviewing patients to identify people at risk.
“The challenge is that, right now, it takes a significant amount of time to identify close contacts of the person who has been infected,” says Akshay Saigal, Head of Innovation Labs for Asia at DXC Technology. “The track and trace is very manual and takes huge teams of dedicated officials to perform.”
He thinks that tech innovation can make a big difference to nations in need and spoke with GovInsider about three steps that officials can use for faster contact tracing.
1. Video analytics to find those at risk
The first step in contact tracing is to identify whom a patient recently met with. Currently, officials must interview patients and try to use other data such as CCTV footage from places they visited, Saigal says.
Even before tracing begins, identifying the infected patient in the video feed can be a tough task. Officials have to rely on details like what clothes the patient was wearing, or which entrance they entered a building from, to spot the patient amidst the busy crowds.
Teaching AI to trace body movements and body signatures can make a huge difference, Saigal notes. “A system can track whether someone has sneezed or coughed, zooming in on them to identify a potential patient. The system can leverage Body API’s – tracking movements with AI – to measure various movements and patterns,” he says. AI learns quickly.
How can officials track this person when they wear different clothes or aren’t always visible? The way a person walks is helpful. With additional hardware capabilities in cameras, “my gait is a digital signature of my identity,” says Saigal. “Gait generally doesn’t vary much during different conditions, such as with pregnancy or injuries,” he points out.
Facial features also help, even when people wear masks. “It’s possible to extract various facial features like the colour of your skin, and if you’re not wearing glasses, your eyes. We can also pick up various other body identifiers like moles and tattoos,” says Saigal.
AI can even predict what someone is likely to have worn on a said day, helping identify them. “If I’m wearing white or grey for three or four days, on the fifth day or sixth day, I will go with the light colours only. You won’t see me in a dark purple shirt,” he explains. “That’s human psychology based on the age and on the work that they’re doing,” he adds.
2. Tracing the network of infected individuals with graph analysis
Contact tracers need to combine data from a large number of different sources. In today’s world, a patient’s digital identity can be recorded using the national identification number; mobile number; IMEI number, which is a unique serial number tagged to all mobile phones; various social platform handles; biometrics such as facial features and gait; ride-sharing apps and e-payments. A graph database can help make the identification process less manual, Saigal notes.
For one, it can easily map out a patient’s network of interactions, during a person’s journey, showing at risk clusters and helping understand the spread of the virus.
A graph database also integrates multiple sources of data to expand its understanding of the situation. It is difficult to rely on any one identification method during contact tracing – rather, a combination of various input sources will help officials track down the right person more accurately and quickly.
A government could combine various video analytics with mobile phone signal data to understand all of the people in a situation who were at risk, say a busy metro station. “The more dimensions I have, the more correlations I can make. This helps to narrow down the right person to be quarantined,” he adds.
3. Helping officials to prioritise
Who is most at risk, and needs to be reached soonest? How can officials prioritise and adapt to a changing situation?
AI can give a confidence score of how accurate it thinks its predictions are. “This is about how confident the AI is that the identified person is the person of interest,” says Saigal. Government officials can then decide the margin of error they want to work with, and whether they would like to look more closely at someone who is, say, only 85 per cent similar to the person of interest.
The intrusiveness of contact tracing may raise privacy concerns, but these can be addressed by having clear guidelines for who can access the data, and how long the data is kept. For instance, GovTech Singapore’s latest contact tracing app, TraceTogether, uses Bluetooth connections to trace movement, and officials can only access a user’s data if they have tested positive for COVID-19. This is an effective data collection process if personal data protection legislation allows, as most patients would be willing to help others fight this pandemic.
DXC uses cloud computing and machine learning algorithms, such as facial recognition and entity resolution analysis, to track persons of interest across various media (video, mobile signal data) and social platforms. The IT services company also makes use of leading cloud platforms for their image recognition and deep learning capabilities.
In a public health emergency, keeping track of a disease’s spread is crucial to stopping it. New tools, techniques, and innovation can help the tireless contact tracers identify people at risk, save lives faster, and stop a disease in its tracks.