Samsung’s U.S. radiology division has tapped the South Korean artificial intelligence developer Lunit to supply its chest X-ray scanners with programs that automatically highlight critical conditions.
The electronics giant signed a contract through its Boston Imaging unit covering the U.S. market, for its GM85 and GC85A Vision devices. The three-year agreement also includes hardware in Canada and Europe, and has plans to expand to the Middle East, South America and Southeast Asia in the future, according to Lunit.
The deal is built around two AI offerings: Insight CXR and Insight CXR Triage, which together can detect 10 of the most common lung abnormalities—including cancer, pneumonia, pleural effusion and a collapsed lung—and then flag the patient’s images for priority review and expedited treatment.
“Chest X-rays are still today the most requested exam in radiography,” Eric Sumner, head of digital radiography at Boston Imaging, said in a statement. “As part of our voice of customer initiatives, we are building on our prior offering of auto lung nodule detection by integrating new on-device, real-time artificial intelligence technologies.”
The tie-up also features a development collaboration, where Samsung will assist in validating AI solutions for conditions such as an enlarged heart muscle, calcification, fibrosis and tuberculosis.
“We’re excited about the potential this partnership holds for advancing chest X-ray practices, particularly in ICUs and ERs,” said Lunit CEO Brandon Suh. The AI developer has previously signed supply deals for Insight CXR with GE HealthCare.
Lunit recently published findings from a head-to-head study in the RSNA journal Radiology, showing that its Insight CXR program could outperform human readers at spotting lung nodules suspected of harboring cancer.
A team of 17 radiologists and radiology residents, led by researchers at Radboud University Medical Center in the Netherlands, examined scans taken from 386 patients—including 144 confirmed to have at least one nodule—and delivered average sensitivity and specificity ratings of 71% and 80%, respectively. Insight CXR, meanwhile, was able to post a sensitivity of 89%, while maintaining the 80% in specificity.