The novel coronavirus pandemic has been managed with widely varying degrees of success around the world. Artificial intelligence (AI), which can help to power all sorts of efforts, has been enlisted thus far in limited ways. But researchers at a virtual conference held on April 1 by the Stanford Institute for Human-Centered Artificial Intelligence explored some of the ongoing and potential applications of AI to systematize efforts to fight COVID-19.
These include automated telehealth systems, apps for tracking and tracing exposed individuals, wearables and connected devices to monitor for COVID-19 symptoms and disease progression, and at-home testing for the novel coronavirus. AI is even being tapped in the race to explore and identify the most promising treatment and vaccine candidates.
Learning from Asia
Some of the most effective implementations of COVID-19 control measures have been seen in many of the Asian countries that have experience with prior epidemics, most recently with SARS in 2003. Singapore and Taiwan were both highlighted by researchers for their sophisticated use of mobile phones to quarantine and monitor incoming travelers as well as the broader population.
Singapore has been tracking cell phones via GPS to ensure quarantine of infected or exposed individuals. It also developed a track-and-trace app known as Trace Together that has been downloaded by more than 600,000 people and is now freely available.
“Cellphones with this app communicated with Bluetooth networks and enabled contact tracing for 21 days. And, actually these apps could calculate the distance between people, so you knew if you were coming close to a COVID-19 person,” explained Michele Barry, a professor and senior associate dean of global health and director of Stanford’s Center for Innovation in Global Health.
Massive track-and-trace efforts of travelers and residents facilitated by technology, as well as widely available testing for the novel coronavirus and systematic temperature monitoring in many public places have been key in some of the most successful efforts in Asia. The latter effort was often conducted by hand, but also via camera-based thermosensor systems as well as temperature screening drones.
In Taiwan, based on cellphone location tracking data the government texted to offer testing and masks to potentially exposed people who had been in the area of known COVID-19 cases. China, South Korea, Israel and various other countries have developed comparable apps to alert individuals if they have come into contact with an infected person, as well as to track quarantine compliance.
Privacy and civil liberties raise issues in Western countries around some of these top-down, technology-enabled government efforts. Track-and-trace apps, and the underlying AI to power them, are starting to roll-out in the U.S – but these rely more on crowd-sourced, anonymized data that is intended to make these tools less invasive and better protect individual information.
Stanford mechanical engineering PhD candidate Tina White presented at the virtual conference on her work on a crowd-sourced app to track COVID-19 in the U.S. She noted that privacy regulations in the U.S. would prohibit connecting GPS data, which is very identifying, with medical data.
Rather than GPS, this approach would rely on Bluetooth data that would more easily allow for de-identified individual data. The technology also implies proximity of the cellphones themselves.
In recent days, more emphasis has been placed on the role of asymptomatic patients in the novel coronavirus epidemic. “The intervention that we're proposing captures some of these events where you have asymptomatic transmission,” said White. “If people would voluntarily put into their phone that they were a confirmed case, then the people that they contacted are anonymously alerted. Those people who were contacted can be targeted with intervention, specifically asked to stay home for the next few weeks.”
Separately, Mount Sinai Health System has just launched an app targeted to New York City that is based on daily survey data. But this is designed to better inform public health preparedness efforts, rather than individual behaviors.
Sensors, wearables and telehealth
Another approach to identifying asymptomatic people, as well as to help protect some of the most vulnerable groups such as the elderly, is systematic distribution of monitors and wearables that track physical metrics such as temperature and heart rate.
“A potential opportunity is to use AI-powered smart sensor technology. The idea here is that sensors installed at home can help families and clinicians keep track of the health conditions of the elderly in versatile and scalable fashion,” explained Fei-Fei Li, the co-director of the Stanford Institute for Human-Centered Artificial Intelligence.
“Here's how the system works in a four-step overview. First is collecting data by putting sensors at home,” continued Li. “Camera sensors carry a lot of detailed information of a person's activity, but it's the least compatible with most people's privacy needs. Other sensors include depth sensors, thermal sensors, as well as wearable sensors. in the last color.”
Another researcher pointed to other wearable efforts as potentially useful, particularly for identifying asymptomatic cases of COVID-19 in the absence of widespread antibody testing.
“Being able to identify even asymptomatic people is pretty critical. There's some interesting sort of studies out from companies like Oura, the smart ring, and Whoop, the smart band,” said John Brownstein, a pediatrics professor at Harvard Medical School. “Both of those show some early evidence that you can actually detect those physiological changes even if a person isn’t experiencing symptoms.”
“There are some interesting opportunities around wearables and continuous tracking that ultimately can identify cases, even with those who don't actually experience symptoms,” he added. “They're identifying cases days ahead of someone actually experiencing symptoms. They are deploying these in certain populations with some really exciting results.”
Finland-based Oura Health Ltd., which also has a San Francisco office, and the University of California at San Francisco have started a COVID-19 symptom monitoring clinical trial in 2,000 health care workers, as well as in its existing users. The ring tracks body temperature, heart rate and respiratory rate. Whoop has a wristword fitness band that monitors respiratory rate; it is starting a COVID-19 symptom study in conjunction with the Cleveland Clinic.
Beyond tracking COVID-19 in the population and individuals, the pandemic is also placing an emphasis on the need for telehealth to better assess and manage disease, as well as at-home infection testing –both of which can be backed by AI.
“Any form of telehealth at scale is a really key, whether you have symptoms, or you don't,” said co-founder and chief technology officer of Curai, Xavier Amatriain. “If you have a question and you need to talk to a provider or a doctor, you want to do it from your home, and you don't want people to leave their house, even if they want to get tested. That's why at home testing becomes so important.” The Palo Alto, Calif.-based startup specializes in AI and machine learning-enabled telehealth services.
“We only want people to leave their house and go to hospital in a really critical situation. I think the symptom checkers and all these other automated assessment tools are important enough for them to build up with the next steps,” he concluded. “How do we keep people from leaving their houses –without incurring any risk for themselves and others?”