Nvidia Clara Federated Learning Introduces New AI Learning Technique to Healthcare SectorDecember 5, 2019
Nvida Clara Federated Learning is a distributed, collaborative technique for AI model training that keeps patient data where it belongs — inside the walls of a healthcare provider.
The Clara FL runs on Nvidia EGX intelligent computing platforms that can perform local deep learning training and collaborative training for a more accurate global model. Its collaborative training is done by the Nvidia EGX platform which securely provisions the federated server and collaborating clients, delivering everything required to initiate a federated learning project, including application containers and its initial AI model. Hospitals will then be able to label their own patient data using Nvidia Clara AI-Assisted Annotation SDK that is integrated into medical viewers like 3D slicer, MITK, Fovia and Philips Intellispace Discovery. Nvidia AI helps radiologists in labeling and reduces the time needed for complex 3D studies.
To train AI models and reach their desired accuracy, NVIDIA EGX servers at participating hospitals train the global model with their local data. The training results are then shared with the federated learning server via a secure link that prioritizes patient data privacy.
The training results shared with the federated learning server are partial model weights and have no patient records. This distributed approach delivers exceptional performance in deep learning while preserving patient data privacy.