 |
|
The XPRIZE team included, from left: Dan Piatkowski, PhD, Alison Freifeld, MD, Fadi Alsaleem, PhD, and Basheer Qolomany, PhD (Photos by Craig Chandler/UNL, Kent Sievers/UNMC, Ryan Soderlin/UNO and Erika Pritchard/UNK) |
A team of researchers from three University of Nebraska institutions — including UNMC’s Alison Freifeld, MD — finished among the Top 10 teams in a global competition to develop an artificial intelligence-driven model to advise policymakers on how best to handle the COVID-19 pandemic.
In December, after Phase 1 of the $500K Pandemic Response Challenge — run by XPRIZE, which designs and operates incentive competitions to solve the world’s grand challenges, and sponsored by Cognizant — the Nebraska team was among the 48 teams from 17 countries chosen to advance to the final phase. Nearly 500 teams from around the world entered this competition.
Phase 1
In Phase 1 of the competition, teams were tasked with analyzing local COVID-19 data, intervention strategies and mitigation policies to develop and test a prediction model that could anticipate global infection spikes.
“Our model is good at predicting future cases, but when you think about the state government, what can you do to reduce cases in Nebraska? That’s why we need a predictive model,” Dr. Alsaleem said. “If you close the schools, we expect the numbers to do this. If people are asked to work from home, the numbers do that. Phase 2 is about optimization of the response, giving government a way to evaluate various scenarios and all the factors to come up with the best plan that can have minimal impact on people’s lives.”
Despite the rush, Dr. Alsaleem said, the Nebraska entry was among the most accurate models.
“We arrived late in the game, two weeks from deadline. We hurried to put something together and our goal was to be among the top 50. Things change after that.
“Most of the work and the data is focusing on what is happening here in Nebraska, but to apply the model across the whole world and see that it’s performing at least as well as other models is very gratifying. It was a good surprise.”
That original project was the starting point for the work done in the Pandemic Response Challenge. After learning about the XPRIZE competition this past fall, the Nebraska team had to work quickly to enter by the November deadline.
Fadi Alsaleem, PhD, an expert in Big Data analysis and assistant professor of architectural engineering at the University of Nebraska-Lincoln, put together a team nearly a year ago to tackle another pandemic-related project in which data gathered from Kinsa thermometers equipped with Bluetooth technology was gathered from various communities in Nebraska and was used to help predict outbreaks across the state.
The Nebraska team includes:
- Dr. Alsaleem, assistant professor of architectural engineering at the University of Nebraska-Lincoln and an expert in Big Data;
- Dr. Freifeld, professor of internal medicine at UNMC and an infectious diseases expert;
- Basheer Qolomany, PhD, assistant professor of computer science at the University of Nebraska-Kearney and an expert in deep learning in support of smart services and disease progression;
- Dan Piatkowski, PhD, assistant professor community and regional planning in the College of Architecture at the University of Nebraska-Lincoln; and
- Three of Dr. Alsaleem’s graduate students — Ali Hazem Al Ramini, Mostafa Rafaie and Mohammad Ali Takallou. Currently, Rafaie is a data scientist with Mutual of Omaha.
The Pandemic Response Challenge aims to harness the power of data and artificial intelligence in equipping policymakers, health officials, and business leaders with insights and guidance necessary to implement public safety measures and safely deliver the vaccine, maximizing their ability to keep local economies open while minimizing potential virus breakouts.
Phase 2 of the contest, which had a deadline of Feb. 3, 2021, required the finalists to develop a prescriptor model — or prescribed action plan — building on their successful entry in Phase 1 (see sidebar). These models were evaluated against key benchmarks, including minimizing the number of cases and minimizing the stringency (i.e. cost) of intervention plans.
Instead of focusing solely on data gathered in Nebraska, the team was required to study the results of responses from around the world when developing its model. That, Dr. Alsaleem said, caused a little concern.
|
|
“Taking a global approach is encouraging a special regime. You do it for one place and it works, but sometimes you model something else and it fails,” Dr. Alsaleem said
Even with limited data, Dr. Qolomany said, the predictions were surprisingly accurate and well-behaved.
“With deep learning types of models, as more data becomes available, the approach can be increasingly useful in dealing with COVID-19 as well as possible future pandemics,”
said Dr. Qolomany. “In this project, the AI system is not designed to replace human decision-makers, but instead to empower them: Humans choose which tradeoffs are the best, and the AI makes suggestions on how they can be achieved.”
The medical and clinical perspectives Dr. Freifeld provided allowed for “a bird’s-eye view of the pandemic” that she said helped the team determine which parameters were most important when developing the models.
“We’re all in our own bubbles of sorts when it comes to science. We have our own areas of expertise, and the point of this challenge is to have a breadth of knowledge in the team,” Dr. Freifeld said. “It’s such a privilege to work with colleagues across the University of Nebraska system on a project of such global importance. This is a very rare occurrence and it should be supported and celebrated.”
Allison,
Way to go!!
Carol Russell
Proud of your accomplishments, Allison!