How technology and AI can help in a pandemic

Autore : Alessandro Piol

Data: 22-04-2020

Tipo: Other

Tematica: Innovation

Events like the current COVID-19 pandemic give us a sense of how important science and technology can be in mitigating and preventing a crisis. While Artificial Intelligence (AI) has been a controversial topic in the past few years, this is a situation where the use of technology and AI can make a difference, by helping predict and flag emerging epidemics, manage the spread of the virus, and speed up the discovery of drugs and vaccines.


Predicting a pandemic


Predicting the arrival of a pandemic is no easy task, but in an interconnected world, where new information is constantly published online and in social networks, and anyone is just an email away, it is today possible to get early warnings that something bad is about to happen.


One example is HealthMap, an automated system created at Boston Children’s Hospital. The system, which scans online news and social media reports from all over the world, including Chinese language sources like WeChat and Weibo, gathered enough proof to send out an alert about unidentified pneumonia cases in the Chinese city of Wuhan. It was the first outside of China to issue a public alert the night of December 30th, 2019.


Another system, ProMED (Program for Monitoring Emerging Diseases) is part of the International Society for Infectious Diseases (ISID). ProMED manages a multidisciplinary global team of over 50 subject matter experts, dedicated to preserving the health of the planet, and reporting from 34 different countries. The information, in this case, is not collected automatically from the internet, but it’s collected by the remote experts who can then feed it back to moderators and editors. At about the same time as the first alert by HealthMap, ProMed was about to publish its own report after some of its experts gathered information from their contacts in China, alerting them to look into “unexplained cases of pneumonia” in Wuhan.


HealthMap and ProMed are two of the leading early warning systems for the detection of infectious disease outbreaks and help inform global agencies such as the World Health Organization, giving experts an early start when bureaucratic obstacles and language differences might slow the process down. Some systems, like ProMed, rely on human expertise. Others are partly or completely automated, like HealthMap. In predicting pandemics, the combination of the two seems to be a good solution.


Processing large amounts of data is best done through computers, but not everything can be done by “web scraping” and then applying natural language processing and machine learning algorithms to the large amounts of data being collected. That’s because the algorithms can only be as effective as the data that has been collected, and sometimes the data could be inconsistent or ambiguous. In many cases, the best solution requires a “human in the middle.” The output of the artificial intelligence engine is analyzed by an expert who can read the signals, but also spot anomalies and inconsistencies. If necessary, the expert can “drill down” and get to the underlying data to understand how the system arrived at its conclusions.


Managing the Spread


In a pandemic, being able to predict the movement and presence of infected people allows decision-makers to make early informed recommendations and policy decisions. Early intervention is among the most effective strategies for treating the spread of contagious diseases. Imagine knowing early on who might have been in contact with an infected individual, and being able to quarantine that subset of the population, as opposed to a whole city, region or country.


In early January, a team at Toronto General Hospital analyzed commercial flight data to see which cities outside mainland China were most connected to Wuhan. Wuhan stopped outbound commercial air travel in late January, but by then an estimated 5 million people had fled the city. The highest volume of flights from Wuhan was to Thailand, Japan, and Hong Kong, and that’s exactly where the virus became noticeable next. (A similar approach was used in 2016 to predict the spread of the Zika virus from Brazil to southern Florida.)


When an infectious disease is detected, public health officials use contact tracing as their first line of defense against its spread. This is a technique that’s been around for decades, and which typically entails health care workers contacting people by phone — a process that can take hours or even days. Contact tracing is effective because it disrupts the chain of transmission. It gives people the opportunity to isolate themselves before infecting others and to seek treatment before they present symptoms.


Technology can make this process much more efficient and quick. On April 10th, Apple and Google announced an (unprecedented) joint initiative to deploy contact tracing functionality to the billions of devices running iOS or Android. The proposed system could identify every recent contact with an infected person, and notify all of them within moments of a confirmed, positive diagnosis.


Unlike a similar system deployed in China, the Google-Apple system is built to be completely anonymous, protecting the user’s privacy. Device identifiers are decoupled from the identity of the owner, and there is no need to use GPS technology and collect location data: Bluetooth happens to have a transmission range similar to the physical proximity required for the airborne transmission of infectious disease. Hence it can tell if someone has been within the range of an infected person. Bluetooth is also on every consumer smartphone and has good power efficiency, so it won’t drain your battery.


Discovering Drugs and Vaccines


AI can also be instrumental in the discovery of new drugs and vaccines, and help researchers navigate the vast ocean of information that’s available for the life sciences and build a usable knowledge base.


Biomedical knowledge is the key to accelerate discoveries and identify hypotheses leading to novel diagnostic tools, therapies, and vaccines. Biomedical researchers struggle with information overload while attempting to grapple with the vast and rapidly expanding base of biomedical knowledge including documents (e.g., papers, patents, clinical trials) and databases (e.g., genes, proteins, pathways, drugs, diseases, medical terms). This is a major pain point for researchers, and with no appropriate solution available they are forced to use search tools (PubMed and Google Scholar) and cumbersome exploration of manually-curated databases. These tools are suitable for finding documents matching keywords (e.g., a single gene or a published journal paper), but not for acquiring comprehensive knowledge about a topic area or subdomain (e.g., COVID-19), or for interpreting the results of high throughput biology experiments such as gene sequencing, protein expression, or screening chemical compounds. To acquire this kind of knowledge is a cumbersome and time-consuming task that can take days or weeks.


There are a few companies trying to solve this problem. Some are using AI as part of a proprietary solution leading to drug discovery. A British company called Benevolent AI is such an example. The objective is to use AI to help formulate a hypothesis that predicts the underlying cause of disease and follows that through many rounds of experimentation and testing until the hypothesis is validated. That leads to the process of designing, synthesizing and developing a new compound to treat that specific disease.


Other companies are developing AI-driven platforms to help biomedical researchers in pharmaceutical companies, academia, and other research institutions, to make sense of all the information available. One such example is Epistemic AI, that provides researchers with the tools to assemble, link and explore all the information that may be relevant to validating a hypothesis. They ingest structured and unstructured data from a variety of sources and databases and use natural language processing and machine learning algorithms to provide the researcher with a fine-tuned set of results. The researcher can then interact with the system (another example of “human in the middle”) to select what’s important and the system will refine the output accordingly. The end result is a knowledge map that can, for example, help find links between the genetic and biological properties of diseases and the composition and action of drugs. Today researchers could use this knowledge to understand what other viruses are similar to COVID-19, see how they work and then figure out whether there are any drugs that could be used to inhibit the virus. Just last week, the company launched a free portal to do just that: it lets researchers extract knowledge from a number of databases including CORD, a COVID-19-specific collection of research papers.


These types of knowledge discovery platforms can increase the efficiency of the research process, by making all the necessary information available in an integrated fashion, and reducing the time of knowledge acquisitions from weeks to hours. It is yet another example of how AI can have an impact on the current pandemic by helping find a cure faster.


Let me remind you that credit is the lifeblood of business, the lifeblood of prices and jobs.

Herbert Hoover […]


Good health is essential to social and economical development and it empowers all of the public sectors.

World Health Organization […]


Innovation is the specific instrument of entrepreneurship. The act that endows resources with a new capacity to create wealth.

Peter Drucker […]

Human Capital

Talent is a source from which water flows constantly renewed. But this source loses its value unless it is properly used.

Ludwig Wittgenstein […]