Graphics processing unit technology developer Nvidia has introduced Nvidia Clara Federated Learning (Clara FL) at this week's RSNA 2019 meeting in Chicago. The application addresses the concerns about security around patient data needed to train artificial intelligence (AI) models.
Running on the Nvidia EGX intelligent edge computing platform, Clara FL facilitates distributed, collaborative AI model training without the need for sharing personal data. The Nvidia EGX platform safely provisions the federated server and collaborating clients, and components such as application containers and an initial AI model can be delivered to initiate a federated learning project. Hospitals train the global model on local datasets, with results being shared with the federated learning server via a secure link. By sharing only partial model weights and not patient records, privacy is preserved, yet the process of deep learning persists until the AI model achieves desired accuracy, according to Nvidia.
Clara FL has found international reach with hospitals in the global healthcare community employing it, such as Peking University First Hospital, Wexner Medical Center at Ohio State University, and King's College London, to name a few, and healthcare organizations such as the American College of Radiology.
