GE Healthcare Captures FDA Clearance for Deep-Learning MRI Reconstruction

The U.S. Food & Drug Administration (FDA) has provided 510(k) clearance to GE Healthcare’s AIR Recon DL, a tool that uses a deep-learning neural network to create shorter scan times while increasing image quality across anatomies.

According to a company statement, this tool, developed on GE’s Edison intelligence platform, integrates into the clinical workflow to generate AIR Recon DL images in real-time at the operator’s console.

“In our experience [AIR Recon DL] enables us to back off on the number of [signal] averages or achieve a higher matrix, to either save on scan time or achieve a higher resolution imaging,” said Hollis Potter, M.D., chairman of the radiology and imaging department at Hospital for Special Surgery in New York City.

Based on company details, AIR Recon DL uses raw data to capture maximum image quality. Alongside improving the signal-to-noise ratio, the tool offers intelligent ringing suppression that is intended to preserve fine image details. This feature aims to address image noise and ringing – two common complaints for radiologists and technologists.

The tool was developed with international partners, and it was evaluated on thousands of cases from a range of anatomies and patient demographics, company officials said. It is available on GE Healthcare’s 3.0T MRI systems and can be acquired as an upgrade or with new system purchases.

Clinical feedback highlighted sharper, less noisy images that allowed for shorter scan times, improved reader confidence, less need for repeat scans, and more scan-to-scan consistency.

“AIR Recon DL benefits clinicians, technologists, and patients alike,” said Jie Xue, president and chief executive officer of GE Healthcare MR. “As we transition to a post-COVID world, MR providers face a significant backlog of patient exams. AIR Recon DL can not only help providers scan more patients per day, but also allows more time to disinfect equipment between patients.”

GE’s global partners in development and clinical validation were: Hospital for Special Surgery, University of California San Francicso, RadNet, University of Wisconsin-Madison, MD Anderson Cancer Center, Medical College of Wisconsin, Centre Cardiologique due Nord in France, Erasmus Medical Center in the Netherlands, Centro Cardiologico Monzino in Italy, University of Yamanashi and Keio University in Japan, and Asan Medical Cetner and Haeundae Paik Hospital in Korea.