Graphics processing unit technology provider Nvidia and King's College London in the U.K. have launched Medical Open Network for AI (MONAI), an open-source framework created specifically to support the scientific development of medical imaging artificial intelligence (AI).
MONAI is designed to handle formats, resolutions, and metainformation of medical images, according to Nvidia. The first public release of the framework provides domain-specific data transforms, neural network architectures, and evaluation methods for measuring the quality of medical imaging models, Nvidia said.
The open-source code is based on the Ignite and PyTorch deep-learning frameworks and integrates libraries for tasks such as data processing, 2D classification, and 3D segmentation. NVIDIA said that MONAI can easily be added by researchers to their existing code; its customizable design enables modular components to be integrated into AI workflows.
"Researchers need a flexible, powerful, and composable framework that allows them to do innovative medical AI research, while providing the robustness, testing, and documentation necessary for safe hospital deployment," said Jorge Cardoso, PhD, chief technology officer of the London Medical Imaging & AI Centre for Value-based Healthcare, in a blog post from Nvidia. "Such a tool was missing prior to Project MONAI."
Nvidia said that entry-level researchers can use MONAI to define an end-to-end AI training workflow with help from its tutorials and application programming interface. MONAI also can enable reproducibility of AI experiments, according to the company.
"Project MONAI is providing a framework by which AI development for medical imaging can be validated and refined by the community with data and techniques from the world over," said Jayashree Kalpathy-Cramer, PhD, of the Massachusetts General Hospital (MGH) & Brigham and Women's Hospital (BWH) Center for Clinical Data Science, in the Nvidia blog post.
Project MONAI was led by Nvidia and King's College London in collaboration with academic advisory board members from the Chinese Academy of Sciences, the German Cancer Research Center, the MGH & BWH Center for Clinical Data Science, Stanford University, and the Technical University of Munich.
MONAI is available on GitHub. Nvidia noted that future releases of its Clara medical imaging tools will also leverage the framework.
