Nvidia unveiled a new federated learning edge computing reference application for radiology to help hospitals crunch medical data for better disease detection while protecting patient privacy.
Called Clara Federal Learning, the system relies on Nvidia EGX, a computing platform which was announced earlier in 2019. It uses the Jetson Nano low wattage computer which can provide up to one-half trillion operations per second of processing for tasks like image recognition. EGX allows low-latency artificial intelligence at the edge to act on data, in this case images from MRIs, CT scans and more.
Nvidia made its announcement of Clara on Sunday at the Radiological Society of North America conference in Chicago. Hospitals will be able to use edge servers from various manufacturers to allow client's systems, like a new portable MRI machine from Hyperfine, to perform deep learning training locally and collaborate with other hospitals for training on a more accurate global model.
Companies including Lenovo, Dell, Cisco and Hewlett Packard Enterprise are providing edge servers for the system.
An EGX server in a hospital can train the global model on their local data and the local training results are then shared to a federated learning server over a secure link without the need to transfer private patient data. The entire process is expected to help initial diagnosis of diseases such as brain tumors and a wide range of diseases.
Four London teaching hospitals will use the process initially, which will expand to more than a dozen UK hospitals in 2020, Nvidia said in a blog.
Those AI services will focus on cancer, heart failure and neurodegenerative diagnoses.
In the U.S., several healthcare providers are involved in the project, including the American College of Radiology, Massachusetts General Hospital and UCLA Medical Center
Nvidia also announced Clara AGX, an AI developer kit to process image and video at high data rates. AGX is powered by Xavier SoCs from Nvidia, the same processors used for self-driving cars. Since they only use 10 watts of power, they can be embedded inside a medical instrument.
The Hyperfine mobile MRI machine uses Clara AGX to allow image processing combined with AI to create anatomical images inside the actual device, said Kimberly Powell, general manager of healthcare at Nvidia, in comments to reporters.
A Clara software developer kit will be available in second quarter 2020 with reference applications for AI inference on streaming endoscopy video and software beamforming for ultrasound.
Powell said Clara was developed over two years with the involvement of thousands of developers. With so much data generated by radiology departments at hospitals the demand for edge computing is more important, she said.