Monai open source AI imaging framework is adopted by cancer research center of Germany, King’s College London, Massachusetts General Hospital, NVIDIA, Stanford University and Vanderbilt University.
Monai (medical open network for AI) is an open source framework optimized for healthcare. NVIDIA Clara application framework, which will be released soon, has been put into production for AI applications in the fields of health care and life sciences.
Monai was launched in April and has been adopted by leading healthcare research institutions. It is a framework based on pytorch, which can be used in the development of medical imaging through industry-specific data processing, high-performance training workflow and advanced repeatable reference implementation.
As part of the updated Clara product, Monai will offer more than 20 pre training models, including the model recently developed for the cowid-19 and the latest training optimization for NVIDIA dgxa100gpu. The optimization can increase the training speed by six times.
Massachusetts General Hospital AthinoulaA.MarTInos Dr. jayashree kalpathy Cramer, director of the qtim laboratory at the biomedical imaging center, said: “Monai is becoming a pytorch in the field of health care, paving the way for closer cooperation between data scientists and clinicians. With the help of federal learning, the adoption of Monai worldwide promotes global cooperation. “
Monai is widely used in medical ecosystem. The German Cancer Research Center, King’s College London, Massachusetts General Hospital, Stanford University and Vanderberg university are all using this AI imaging framework. Monai has been used in everything from the industry-leading imaging competition to the first novice training camp for the framework held in September. The camp attracted more than 550 participants from 40 countries, including undergraduates.
Dr Bennett Landman of Vanderbilt University said: “Monai has rapidly become the preferred deep learning framework in the field of health care. This step from research to production is very important for AI application in clinical nursing. NVIDIA is committed to community driven scientific research, enabling academia to contribute to a framework that can be used for production. This will help further innovation to build enterprise ready features. “
NVIDIA Clara brings the latest breakthroughs in AI assisted annotation, federated learning, and production deployment to the Monai community.
Its latest version has revolutionized AI assisted annotation, allowing radiologists to use a new model called deepgrow3d. In this way, complex 3DCT data can be marked with only one tenth of the original number of hits. The traditional method of segmenting organs or lesions by image or slice is very time-consuming. For large organs such as liver, it may need 250 clicks at most. Today, users can segment with fewer clicks.
NVIDIA Clara’s AI assisted annotation tool and new deepgrow3d function can be combined with foviaai’s f.a.s.t.ai annotation software to mark training data and assist radiologists in film reading. Fovia provides xstreamhdvr SDK suite to view DICOM images through industry-leading PACS viewer.
AI assisted annotation is the key to unlock rich radiology data sets. This technique has also recently been used to label the public covid-19ct dataset published by the National Institutes of health cancer imaging archive. Subsequently, this marker data set was used in MICCAI certified cowid-19 lung CT lesion segmentation challenge.
Clara federal learning enables 20 hospitals around the world to carry out collaborative research in the near future to develop a common AI model for patients with cowid-19. The exam model can predict the oxygen demand of patients with covid-19, which can be obtained through the NGC software registry and is being clinically validated and evaluated in the Mount Sinai health system in New York, the Brazilian Dian ó s Ticos Da America SA, the Cambridge biomedical research center of the National Institutes of health in the United Kingdom and the National Institutes of health in the United States.
Dr Daniel Rubin, Professor of biomedical data science, radiology and medicine at Stanford University, said: “the Monai software framework provides key components for training and evaluating image-based deep learning models, and its open source approach helps foster growing communities and contribute to exciting advances such as federated learning.”
NVIDIA also extends NVIDIA Clara to digital pathology applications. In this area, the existing open source tools can not cope with the huge image size. Clara for early access to pathology contains a reference pipeline for AI application training and deployment.
Jorge Cardoso, chief technology officer, London medical imaging and AI value medical centre, said: “interoperability of health data, model deployment and clinical pathway integration is an increasingly complex and intertwined topic involving domain specific expertise. The Monai project, combined with other parts of the NVIDIA Clara ecosystem, helps to improve patient care and optimize hospital operations. “