Recently, at RSNA (Radiology North America Annual Conference), Nvidia released the latest application progress of its open source medical imaging AI framework MONAI.
Nvidia will provide MONAI Application Package (MAP) to package AI models for easy deployment by medical imaging companies and institutions. Currently, MAP has been adopted by mainstream cloud platforms such as Amazon Cloud, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure.
The official debut of MONAI (Medical Open Network for Artificial Intelligence) was in April 2020, when Nvidia and King’s College London jointly announced the open source alpha version of Project MONAI. MONAI is optimized for the needs of healthcare researchers and can run with deep learning frameworks such as PyTorch and Ignite.
In recent years, medical imaging AI has developed rapidly, and various AI tools have emerged. However, limited by complex workflows and inconsistent development and deployment standards, rapid and large-scale clinical applications have not yet been realized. This is what MONAI and MAP are aiming at Pain points.
At present, MONAI has been downloaded more than 650,000 times, and has been adopted by well-known medical institutions such as Guy’s and St. Thomas’ Hospital and King’s College Hospital.
Nvidia said that the British National Health Service System (NHS) Trust Fund will use MONAI to provide clinical AI applications for stroke, dementia, heart failure, cancer and other diseases.
MAP is “all in one package” and has entered mainstream cloud platforms such as Amazon Cloud
MAP is provided by MONAI Deploy, as an AI model packaging method, which aims to solve the problem of deploying AI models in medical institutions in the past.
Cincinnati Children’s Hospital is creating MAP to deploy an AI model capable of automatically segmenting whole heart volumes in CT images to aid pediatric heart transplant patients.
Dr. Ryan Moore of the hospital said, “If you want to deploy several AI models in the imaging department to help experts identify a dozen different conditions or achieve semi-automated creation of medical imaging reports, it will take a lot of time and resources to create one for each. The model looks for the right hardware and software infrastructure. This has been ‘possible’ but not ‘feasible’ in the past.”
MAP simplifies this process by standardizing how developers build AI models and package them into deployable clinical applications. If developers package an application using the MONAI Deploy application software development kit, hospitals can easily run the application locally or in the cloud. The MAP specification also integrates medical IT standards, such as the medical imaging interoperability standard DICOM.
It is understood that the MONAI Deploy working group is composed of experts from more than a dozen medical imaging institutions, with the goal of supporting AI application developers and clinical and infrastructure platforms that run AI applications.
For developers, MAP can help researchers easily package and test models in clinical settings, thereby accelerating the evolution of AI models. This allows them to gather real-world feedback to refine and improve the AI.
For cloud service providers, support for MAP (designed using cloud-native technology) can help researchers and enterprises adopting MONAI Deploy to run AI applications on their own platforms through container or native application integration.
Currently, cloud platforms that integrate MONAI Deploy and MAP include Amazon Cloud, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure.
Amazon Cloud integrates the MAP interface into the HealthLake image service, enabling clinicians to view, process and segment medical images in real time.
Google Cloud’s medical imaging suite integrates MONAI into its platform, enabling clinicians to deploy AI-assisted annotation tools that help automate manual and repetitive medical image labeling tasks.
Recently, Nvidia also announced cooperation with Microsoft Azure-driven Nuance Precision Imaging Network and Oracle Cloud Infrastructure. Nuance Precision Imaging Network, a cloud platform that provides AI tools and insights to more than 12,000 medical institutions, will be combined with MONAI next.
Oracle has introduced accelerated computing solutions for the medical industry, including MONAI Deploy, into Oracle Cloud Infrastructure. Starting today, developers can use NVIDIA containers on the Oracle Cloud Marketplace to build MAPs through MONAI Deploy.
Additionally, MONAI Deploy is being adopted by the UK’s National Health Service (NHS), Qure.ai, SimBioSys, and the University of California, San Francisco.
Among them, Qure.ai, a member company of the Nvidia Startup Acceleration Program, has developed medical imaging AI models for use cases such as lung cancer, traumatic brain injury, and tuberculosis. exert influence.
SimBioSys, a Chicago-based NVIDIA Startup Accelerator member, built a 3D virtual representation of a patient’s tumor, using MAP for precision medicine AI applications that help predict how a patient will respond to a particular treatment.
UCSF is developing MAPs for several AI models, including applications such as hip fracture detection, liver and brain tumor segmentation, and knee and breast cancer classification.
MONAI Accelerates Large-Scale Clinical Deployment of Medical Imaging AI
On RSNA, Nvidia announced a partnership with the UK National Health Service (NHS), which will use the AIDE platform built on MONAI to provide AI disease detection tools for medical professionals.
The British public healthcare system consists of 10 NHS trusts, and it is expected that the AIDE platform will be rolled out to 11 NHS hospitals next year, serving 18 million patients by then.
The full name of AIDE is AI Deployment Engine, which is “AI Deployment Engine”. It is jointly built by Nvidia and AI Center for Value Based Healthcare. A consortium led by Eyre and St Thomas’ Hospital NHS Trust.
Nvidia said that AIDE will be open-sourced and released on GitHub on December 7.
According to reports, the combination of MONAI and AIDE can safely and effectively verify, deploy and evaluate medical imaging AI models. These models will be used by the NHS to diagnose and treat diseases such as cancer, stroke and dementia.
The AIDE platform is currently being deployed in four hospitals, including Guy’s and St Thomas’ Hospital, King’s College Hospital, University of East Kent Hospital and University College London Hospitals NHS Trust. Medical professionals at these four institutions serve 5 million patients annually.
James Teo, professor of neurology and data science at King’s College Hospitals NHS, sees the work as exciting, “By deploying this infrastructure of clinical AI tools, we can integrate AI into healthcare delivery. Through these platforms, clinicians can scale The deployment of healthcare AI tools, in turn, helps them make decisions that improve the speed and precision of patient care.”
Currently, the AI Center has developed algorithms that can improve the accuracy of diagnosis for diseases such as COVID-19, breast cancer, brain tumors, stroke and dementia risk.
Researchers, hospitals and start-ups across the healthcare ecosystem are starting to realize the benefits of introducing streamlined AI processes into their work, says Haris Shuaib, AI Centre’s AI Transformation practice lead. MONAI was able to standardize hundreds of AI algorithms, maximizing interoperability and impact, while reducing deployment time from three to six months to just a few weeks.
With AIDE, the AI Center is able to seamlessly and securely link approved AI algorithms to patient records, without the data ever leaving the hospital trust. The analysis results of clinical data will be sent back to the electronic case to help make clinical decisions, which provides a valuable data source for the clinical multidisciplinary team’s disease consultation. The hospital hopes that AIDE can help speed up this process and benefit patients.
“Currently, most AI models have been in the research and development stage, and few of them can be actually used in patient care.” Jorge Cardoso, chief technology officer of the Value-Based Healthcare project at the London Medical Imaging and AI Center, believes that this is where MONAI Deploy plays a role. Help promote the landing of research and development results and realize more influential clinical AI. Leifeng.com Leifeng.com
This article is reproduced from: https://www.leiphone.com/category/healthhai/uOBxOL9w36f7xfNx.html
This site is only for collection, and the copyright belongs to the original author.