Big
data analytics techniques are well-suited for tracking and controlling the
spread of COVID-19 around the world.
By Jessica
Kent
April 02,
2020 - The rapid, global spread of COVID-19 has brought advanced big
data analytics tools front and center, with entities from all sectors of the
healthcare industry seeking to monitor and reduce the impact of this virus.
Researchers and
developers are increasingly using artificial
intelligence, machine learning, and natural language processing to track and
contain coronavirus, as well as gain a more comprehensive understanding of the
disease.
In the months since
COVID-19 hit the US, researchers have been hard at work trying to uncover the
nature of the virus – why it affects some more than others, what measures can
help reduce the spread, and where the disease will likely go next.
At the core of these
efforts is something with which the healthcare industry is very familiar: Data.
“This is, in essence, a big data problem.
We're trying to track the spread of a disease around the world,” James Hendler,
the Tetherless World Professor of Computer, Web, and Cognitive Science at
Rensselaer Polytechnic Institute (RPI) and director of the Rensselaer Institute
for Data Exploration and Applications (IDEA), told HealthITAnalytics.
At RPI, researchers
are using big data and analytics to better comprehend coronavirus from a number
of different angles. The institute recently announced that it would offer
government entities, research organizations, and industry access to innovative
AI tools, as well as experts in data and public health to help combat COVID-19.
“We're working with
several organizations on modeling and dealing with the virus directly using a
supercomputer, and we’ve been creating some websites where we track all the
open data and documents we can find to help our researchers find what they're
looking for,” Hendler said.
“We also have some
work we've been doing in understanding social media responses to the pandemic.
One project in particular has focused on tracking data from Chinese social
media as coronavirus spread there in mid-January, and then comparing it to
American data.”
Between recognizing
signs and symptoms, tracking the virus, and monitoring the availability of
hospital resources, researchers are dealing with enormous amounts of
information – too much for humans to comprehend and analyze on their own. It’s
a situation that is seemingly tailor-made for advanced analytics technologies,
Hendler noted.
“There are several big data components to this
pandemic where artificial intelligence can play a big role,” he said.
“One component is
biomedical research. A lot of work is going on to try to develop a vaccine to
find out whether there are any current drugs work against COVID-19. All of
those projects require molecular modeling, and many of them are using AI and
machine learning to map things we know about the virus to things in pharmacological
databases and genomic databases.”
Several big-name
organizations have launched projects like these – Amazon Web Services, Google Cloud, and others have recently
offered researchers free access to open datasets and analytics tools to help
them develop COVID-19 solutions faster.
“AI can eliminate
many false tracks and allow us to identify potential targets. So instead of
trying 100 or 1000 different things, we can we narrow it down to a much smaller
size much faster. That's going to accelerate the eventual finding of the
vaccine,” Hendler said.
Researchers are also
leveraging AI to evaluate the effects of COVID-19 interventions on individuals
across the country, Hendler stated.
“A second component has to do with natural
language processing and social media. What can we extract from social media
that can help our scientists? What can we learn about how people are bearing
the burdens and stresses of the pandemic?” he said.
“With SARS and other
outbreaks, we never really had to figure out how different social distancing
techniques are impacting the spread in different places. You can't just compare
numbers, because there are a lot of other factors to consider. AI is very good
at that kind of multi-factor learning and a lot of people are trying to apply
those techniques now.”
At UTHealth, a team
developed an AI tool that showed the need for stricter,
immediate interventions in the Greater Houston area. And at Stanford
University, researchers have launched a data-driven model that predicts
possible outcomes of various intervention strategies.
Using big data and
analytics tools of their own, Hendler and his team are aiming to do something
similar.
“We have a lot of
time series data from China, we have information about airline transportation,
and we have population models for each country. Now we’re looking at doing this
in our own region, and seeing if we can track and predict the spread based on
the kind of social measures taken within different regions,” he said.
“We want to prototype
that in our region and then scale it up to the US, and then eventually, the
world.”
AI can also help
organizations draw on research from the past, applying this knowledge to
present and future situations.
“A third area where
AI can make an impact is in mining scientific literature,” Hendler said.
“In past years, you
had hundreds of grad students reading papers and trying to figure out what was
going on. At many universities, there's a lot of effort to say, ‘What can we
learn from what’s already been published?’”
While AI and other
analytics technologies appear to be the best possible tools for assessing and
mitigating a global pandemic, researchers can’t always access what they need to
build these models.
“The ideal data is
hospital data that would tell us who is experiencing certain impacts from the
virus,” Hendler said.
“For example, one
project we'd love to do would be to correlate environmental or genomic factors
to the people who are getting advanced respiratory problems, which is what’s
killing most people with this disease. Is there a genetic component to that? Is
it something where environmental factors are some kind of comorbidity? But can
we can't get that kind of data because of HIPAA restrictions.”
Instead, research
teams should focus on extracting insights from the information they do have
available, Hendler said.
“Information about
how people are moving, the effect of travel restrictions or stay at home
orders, how many people have what – that’s data we can get. The more details we
can get, the better, and a lot of that data is starting to be shared because
you don't have to say who the people are, just where the people are,” he said.
The unprecedented
impact of coronavirus around the world has sparked the need for unprecedented
partnerships, and these collaborations will contribute significantly to finding
viable solutions.
“In healthcare,
academia and industry are mostly set up for people to stay in their own lanes.
But people are rapidly beginning to realize that attacking this problem is
going to require a collaborative effort,” Hendler concluded.
“To make any real
progress in this situation, you need to bring together people who understand
the computation and AI, people who understand the biological and biomedical
implications, and people who understand population models. It's a very
interdisciplinary problem, and to make any headway, we need the data and we
need the team.”
No comments:
Post a Comment