AI in the Fight to Detect and Cure Coronavirus
The COVID-19 coronavirus is rapidly spreading around the globe, threatening communities, alarming financial markets and, in many cases, forcing national epidemic response teams to quickly update and implement years-old plans. As those researchers race to apply today’s technology to detecting and combating the virus, many are talking about how – and even field testing – AI strategies that could be deployed to ameliorate this global crisis.
Tracing the Outbreak
Early evidence of AI’s utility in fighting coronavirus was an AI-enabled early detection of the disease by Toronto startup BlueDot. The company’s AI tool, which is targeted at early detection of potential epidemics, uses machine learning, natural language processing (NLP) and sentiment analysis to process data from new stories, Center for Disease Control (CDC) and World Health Organization (WHO) reports, airline itineraries and more. In late December, BlueDot’s warning system sounded the coronavirus alarm a full week before the CDC and other major health organizations. To learn more about BlueDot’s early detection, visit Doug Black’s article here.
Elsewhere, Boston Children’s Hospital tackled another key facet of controlling the outbreak: mapping it. John Brownstein, the hospital's chief innovation officer and a Harvard professor, told Bloomberg that by applying machine learning and AI to what he calls “digital exhaust” (such as social media posts and search queries), the hospital developed a mapping tool (accessible here) that may be able to detect early events before the news reaches officials.
“Before Chinese New Year we looked at how many people left Wuhan over a day, and this information comes from search engines including Baidu,” explained Moritz Kraemer, a spatial epidemiologist at the University of Oxford who contributed to the tool, in an interview with Lancet Digital Health. “WeChat, a messaging, social media and mobile payment app, provides data on travel around Wuhan. Machine learning models use these data to predict the most likely location of where novel coronavirus might arrive next and this might inform where and how to run border checks.”
According to Brownstein, much of their focus is on making sure there are tools in place to detect outbreaks within the United States. “If we could harness the power of the Internet,” he told Boston 25 News, “we could get a view into an emerging public health threat like coronavirus in a way you never attain through traditional channels like public health agencies collected data.” David Heymann, executive director of the WHO’s Communicable Diseases Cluster, led the organization’s SARS response and agrees with this assessment. “By monitoring social media, newsfeeds or airline ticketing systems, for example,” he told Lancet, “we can tell if there’s something wrong that requires further exploration.”
Diagnosing the Illness
One of the challenges with the coronavirus outbreak on the ground is that COVID-19 detection kits are in short supply and are still experiencing iterations and changes. So when diagnosing COVID-19, radiologists look for a specific expression of pneumonia. While pneumonia isn’t a confirmation unto itself, it helps medical professionals know which patients require more urgent isolation and more thorough diagnostic procedures. With limited test kits, still-evolving knowledge and overworked radiology departments, researchers like Seoul National University Hospital’s Hyungjin Kim argue that AI could help to “alleviate the burden of radiologists and clinicians and enhance rapid triaging.”
In fact, some Chinese hospitals have already deployed AI to help with COVID-19 diagnosis, as reported in a Wired article this week. Infervision, a Beijing startup, helped to deploy an AI-powered COVID-19 pneumonia detection tool at 34 hospitals, helping to examine over 32,000 patients. The tool, which had previously been used mostly to detect cancerous lung nodules, leverages lung images from hospitals to understand and flag lung problems in CT scans. Using images of COVID-19 pneumonia from one of the first Chinese hospitals to treat a patient with the coronavirus, Infervision repurposed the tool for COVID-19 detection. While not yet formally, approved by the Chinese government, Infervision CEO Kuan Chen told Wired that “there are always risks for any actions in a dangerous outbreak like this, but the risk of inaction is much greater.”
Solving the Problem
The long game, of course, is developing effective treatments – or even a vaccine – for COVID-19. Insilico Medicine, a Maryland startup that leverages AI for drug discovery, is one of the companies rallying its resources behind fighting the coronavirus. According to Fortune, within just a few days after using 28 different machine learning models to explore medicinal options, Insilico’s AI tool identified thousands of molecules that may be candidates for COVID-19 drugs. In a statement, Insilico promised to “synthesize and test up to 100 molecules using its own resources and the resources generously offered by its closest partners,” the first six of which have already been completed.
"During this difficult time, every promising approach must be used to expedite the drug discovery efforts against 2019-nCoV, including utilizing the generative chemistry part of our end-to-end drug discovery pipeline,” said Alex Zhavoronkov, CEO of Insilico Medicine. “We encourage medicinal chemists to evaluate the generated molecules, provide their feedback and consider them for synthesis. Our team will also synthesize and test several of the generated compounds.”
Of course, AI can’t do it all – yet. “We can’t replace the human brain at this point, nor the epidemiologist or virologist with anything that can analyze and rapidly do what is necessary at the onset of an outbreak,” Heymann told Lancet. “We still need to prime that AI with information from study of the evidence and link this to events in the outbreak.”
Header image: a close-up of the coronavirus map developed by Boston Children's Hospital and partners, which is accessible here.