Federal AI Diagnostics Research Advanced During COVID-19

Federal AI Diagnostics Research Advanced During COVID-19

The intensive demands of the pandemic have encouraged knowledge sharing and formalized practices of AI for health care research.

The pressures of the COVID-19 pandemic to accelerate medical research has encouraged data sharing and consolidation of practices across federal agencies to lay the groundwork for increasingly sophisticated applications of AI toward diagnostic and health care.

Speaking at the GovernmentCIO Media & Research AI Gov virtual event, Department of Veterans Affairs AI lead Gil Alterovitz and Georgetown University Interim Dean of Research Dr. Moshe Levi discussed how data sharing and refinement, as well as the range of potential applications, have increased markedly over the prior year.

The demands of analyzing a newfound disease’s particular risk factors led to widespread usage of medical data for determining severe COVID-19 comorbidities.

“As COVID started spreading out, it became clear obesity, diabetes and aging were major risk factors," said Levi. "We identified a pathway that is similarly upregulated following COVID-19 infection."

This necessity has also spurred new data applications and sophisticated methodologies across health-focused federal agencies as well, including at VA's National Artificial Intelligence Center (NAIC). This was motivated in large part by the practical necessity of responding to a pandemic that had spread far more rapidly than initially assumed, which has been particularly impactful on VA’s development of “trustworthy AI” — or AI data curated and developed within an ethical and open framework.

“COVID actually made things move faster because it was a use case that really presented an immediate need. So let's say you know surge is happening in a particular location or across multiple locations, there's a need to essentially pilot and leverage is such a system so that you can get things used more quickly,” said Alterovitz, VA's NAIC director.

The refinement and sharing of COVID-relevant AI data has also encouraged collaboration between research labs and clinicians, leading to newfound consolidation of understanding between those departments as well as workflows for enabling future AI development.

“COVID has really enabled people to think in new ways, in the sense that it is really something that I think will carry forward into the future. This kind of process where clinicians and researchers are working together in this iterative way,” Alterovitz said.

Going forward, Alterovitz recommended that federal agencies looking to apply AI toward diagnostics and public health reform their IT systems to better facilitate collaboration between researchers and clinicians, and overall improve knowledge sharing within the organization.

“I think many systems, when you look at them, actually have two different systems. One that's used for the research setting. The supercomputers, the clusters and so forth, where you develop the AI models. Then there's a different system, which is the one that's used clinically and doesn't have the same analytical capabilities. There's kind of that dichotomy. So in the future, there may be a better way to kind of connect these systems,” Alterovitz said.