Repurposed AI Shows Promise for Future Pandemic Response

A new study published in Scientific Reports highlights the potential of repurposing existing AI tools to help combat future pandemics, offering a glimpse into how ready-to-use imaging software might play a critical role in global health emergencies.

During the early stages of the COVID-19 pandemic, it became clear that imaging would be vital not just for diagnosis, but also for gauging disease severity and guiding patient management. While chest CT scans proved highly effective, their availability was limited in many parts of the world. More accessible imaging modalities, such as chest radiographs, became indispensable, particularly in resource-constrained settings.

“With the rapid surge of imaging volume during the pandemic, rapid screening, triaging, and isolation of COVID-19-positive or suspected patients became critical such that precautionary preparations, planning and management can be readily made for hospitals and clinics,” wrote lead author Young Beom Kim of the College of IT Engineering at Pyeongtaek University in the Republic of Korea.

Although several AI tools designed specifically for detecting COVID-19 from imaging were eventually developed, many of these systems emerged too late to meaningfully impact the initial waves of the pandemic. Furthermore, their specificity to COVID data may limit their adaptability in the event of future outbreaks involving different pathogens.

To address this limitation, researchers explored whether AI applications originally developed to detect other lung conditions could be useful during a respiratory pandemic. They tested a commercially available AI software system originally built to detect pulmonary nodules in chest radiographs, applying it to a dataset of X-rays from the early stages of the COVID-19 pandemic. Their goal was to see whether a general-purpose lung imaging AI tool could flag signs of viral pneumonia without any additional training on COVID-specific data.

The results were encouraging. The AI achieved a sensitivity of 86.8% and a specificity of 59.6% in detecting pneumonia features, including those linked to COVID-19. When provided with a single posterior-anterior (PA) chest X-ray, the model performed even better, achieving a sensitivity of 89.2% and a specificity of 67.4%. The model was less accurate with anterior-posterior (AP) views, underlining the importance of standardized imaging protocols in high-pressure scenarios.

The study’s findings suggest that widely available AI systems could be deployed in the early phases of future pandemics, helping frontline providers manage case volume more effectively and rapidly identify high-risk patients—even in settings without disease-specific AI tools on hand.

“This study demonstrated the feasibility and effectiveness of utilizing existing AI technology, originally designed for detecting pulmonary nodules, in addressing urgent healthcare needs during a global pandemic,” the authors wrote. They added that the work provides “compelling evidence for the immediate clinical deployment of existing AI technology in responding to infectious disease outbreaks.”

As health systems look to futureproof their response strategies, such adaptable and scalable AI applications could prove essential in navigating the challenges of future global health crises.