FDA OKs AIRS Medical’s AI-Powered MRI Reconstruction Software SwiftMR

AIRS Medical Inc., a healthcare startup based in South Korea, announced FDA 510(k) clearance of its first product SwiftMR. SwiftMR is an AI-powered MRI reconstruction software that enhances MR images acquired under various conditions, contributing to higher throughput and better patient experience.
 
MRI scans are a safe and reliable source of imaging, but a single procedure takes 20-90 minutes, causing discomfort to patients and sometimes requiring sedation. Hence radiologists are forced to yield image quality to keep scan time at a patient-tolerant level. In addition, long scan times and hiking demand make MRI notorious for wait times that can build up to months.
 
AIRS Medical tackles this dilemma by utilizing deep learning engines trained with high-quality images to enhance low-quality images. Because the software communicates directly and automatically with the institution’s PACS (Picture Archiving and Communication System), it easily boosts performance and efficiency of the entire fleet of MRIs with a quick installation.
 
Since its launch in July, SwiftMR now serves numerous institutions across Korea, with over 4,000 patients scanned and zero downtime to date. The solution is gaining traction fast, as the output images are receiving positive feedback from radiologists in the country and abroad. “Refocusing our goal from increasing quantitative metrics of the deep learning engine to satisfying seasoned radiologists has been critical in advancing the technology to a clinically acceptable level,” said Jingu Lee, CEO of AIRS Medical. “We believe we can meaningfully reduce scan and wait times—and ultimately, enable medical imaging services to be more accessible to people around the world.”
 
AIRS Medical will showcase its market-ready product for the first time in the U.S. at RSNA 2021, where Dr. Jongho Lee, Professor of Seoul National University and scientific advisor of AIRS Medical, will give a presentation on the real-world application of SwiftMR.