Rsna2019 NVIDIA Clara federal learning not only brings AI to hospitals, but also protects patient data

At the annual conference of the Radiological Society of North America (RSNA), more than 100 exhibitors adopted NVIDIA technology to introduce AI into radiology 2019 seems to be the turning point of AI application in the field of medical health Despite the huge potential of AI, there is still a key challenge - how to obtain the massive data needed to train AI models while protecting patients' privacy Working with the industry, NVIDIA has created a solution At this year's RSNA conference, NVIDIA launched NVIDIA Clara federal learning, which uses distributed collaborative learning technology to save patient data in medical service institutions NVIDIA Clara federated learning runs on NVIDIA's recently released NVIDIA egx intelligent edge computing platform Federated learning - the privacy preserving AI Clara federated learning is a reference application for distributed collaborative AI model training, which can protect patients' privacy These distributed client systems run on edge oriented NVIDIA NGC ready servers built by global system manufacturers, which can perform deep learning training locally and cooperate to train more accurate global models How it works: the Clara federated learning application is packaged into the helm diagram to simplify deployment on the kubernetes infrastructure The NVIDIA egx platform will securely configure the Federation server and collaboration client, providing everything needed to launch the federated learning project, including the application container and the initial AI model NVIDIA Clara federal learning uses distributed training to develop AI models across hospitals Hospitals that do not need to share personal data to participate in the project use the NVIDIA Clara AI auxiliary annotation SDK to tag their own patient data The SDK is integrated into medical indicators such as 3dlicense, MITK, fovia and Philips intellispace discovery Using pre trained models and migration learning techniques, NVIDIA AI can help radiologists tag, reducing complex 3D research time from hours to minutes For participating hospitals, their egx servers train the global model based on their local data Local training results are shared back to the federated learning server through a secure link This method can only share the correction of the model weight, but not the cases, so it can protect the privacy and build a new global model through joint averaging This process will be repeated until the AI model reaches its desired accuracy This distributed approach achieves excellent deep learning performance, while ensuring patient data security and privacy British and American medical institutions take the lead in adopting NVIDIA Clara federal learning The world's leading medical and health institutions, including the American College of Radiology (ACR), Massachusetts General Hospital and UCLA Medical Center, are taking the lead in adopting the technology, striving to develop personalized AI applications for their doctors, patients and medical facilities, and their medical number Data, applications, and devices are growing, and patient privacy must be protected ACR is piloting NVIDIA Clara federal learning in its national medical imaging platform, ai-lab AI Lab will help 38000 medical imaging members of ACR safely build, share, adjust and verify AI models For medical institutions that want to use AI Lab, there are many edge oriented nvidiangc ready systems available, including Dell, HPE, Lenovo and supermicro UCLA radiology is also using NVIDIA Clara federal learning to bring AI's powerful capabilities to radiology As a top academic medical center, UCLA is able to verify the effectiveness of Clara's federal learning and expand it into a broader University of California System in the future New England's Partners Healthcare has also announced a new plan to adopt NVIDIA Clara federal learning The clinical data Science Center at Massachusetts General Hospital and Brigham women's hospital will take the lead in this effort, using data assets and clinical expertise from the American alliance healthcare system system In the UK, NVIDIA is working with King's College London and owkin to build a federal learning platform for the national health service The owkinconnect platform running on NVIDIA Clara enables the algorithm to transfer from one hospital to another and train on the local data set of the hospital The platform provides a blockchain distributed ledger for each hospital to capture and track all data used for model training The project initially links four leading teaching hospitals in London to provide AI services to accelerate work in areas such as cancer, heart failure and neurodegenerative diseases, and will expand to at least 12 UK hospitals by 2020 With the rapid popularization of sensors, medical centers like Stanford Hospital are committed to giving intelligence to every system However, to make the sensor intelligent, the device needs a powerful and low-power AI computer This is the reason why NVIDIA wants to release NVIDIA claraagx, an embedded AI development kit that can process images and videos at high data rates, and apply AI reasoning and 3D visualization technology to care points ClaraAGX is equipped with the same NVIDIAXavierSoC as the automatic driving vehicle control processor Its power consumption is only 10W, which is suitable for embedding in medical instruments or running in small adjacent systems The revolutionary hyperfine is the perfect application case of claraagx, which is the world's first portable medical point MRI system Hyperfine will be on display at NVIDIA booth at this week's RSNA conference The hyperfine system is one of the first systems that are expected to use claraagx among many medical instruments, surgical kits, patient monitoring devices and smart medical cameras NVIDIA is witnessing the opening of the AI medical Internet of things NVIDIA Clara SDK will soon be launched through NVIDIA's early access plan It includes two common reference applications - AI reasoning for endoscopic video and software beamforming for ultrasound Jeni turtle