- MONAI is user-friendly, delivers reproducible results and is domain-optimised for the demands of healthcare data
- The open-source code is based on the Ignite and PyTorch deep learning frameworks
NVIDIA along with King’s College London has introduced MONAI (Medical Open Network For AI) an open-source AI framework for healthcare research. The company said that it builds on the best practices from existing tools, including NVIDIA Clara, NiftyNet, DLTK and DeepNeuro.
As per NVIDIA, MONAI is user-friendly, delivers reproducible results and is domain-optimized for the demands of healthcare data. It is equipped to handle the unique formats, resolutions and specialized meta-information of medical images. The company said that the first public release will provide domain-specific data transforms, neural network architectures and evaluation methods to measure the quality of medical imaging models.
Seb Ourselin, head of the School of biomedical engineering & imaging sciences at King’s College London said, “In partnership with NVIDIA, Project MONAI is following industry standards for open-source development and building a global community across academia and industry to establish a high quality framework supporting scientific development in medical imaging AI.”
Ignite and PyTorch deep learning frameworks
NVIDIA and King’s College London are leading the initiative in collaboration with an advisory board hailing from the Chinese Academy of Sciences, the German Cancer Research Center, Kitware, MGH & BWH Center for Clinical Data Science, Stanford University and the Technical University of Munich.
Stephen Aylward, chair of the MONAI advisory board and a senior director at open-source software company Kitware said, “Project MONAI has outstanding potential to accelerate the pace of medical imaging AI research. It provides a high-quality, open-source foundation that is specialized for medical imaging, that welcomes everyone to build upon, and that anyone can use to communicate and compare their ideas.”
The open-source code is based on the Ignite and PyTorch deep learning frameworks. It combines libraries for data processing, 2D classification, 3D segmentation among others. Researchers can bring MONAI to their existing code by using the customisable design to integrate modular components into their AI workflows.
Reproducibility of experiments
Modular, open-source solutions will give researchers the flexibility to customise their deep learning development without the need to replace their existing workflows with an end-to-end system. Jorge Cardoso, chief technology officer of the London Medical Imaging & AI Centre for Value-based Healthcare said, “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. Such a tool was missing prior to Project MONAI.”
The company said that a major goal of the MONAI framework is to enable the reproducibility of experiments. This will make researchers share results and build upon each other’s work to advance the state of the art.
Jayashree Kalpathy-Cramer, scientific director at the MGH & BWH Center for Clinical Data Science said, “Reproducibility of scientific research is of paramount importance, especially when we are talking about the application of AI in medicine. 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.”
