Project MONAI was originally started by NVIDIA and King’s College London to establish an inclusive community of AI researchers for the development and exchange of best practices for AI in healthcare imaging across academia and enterprise researchers.
MONAI Core is the flagship framework created by Project MONAI and it wouldn’t have been possible to accelerate this development without the development of existing toolkits such as NVIDIA Clara Train, NiftyNet, DLTK, and DeepNeuro.
Helping guide MONAI's vision and mission, an Advisory Board and nine working groups, have been created and are led by thought leaders throughout the medical research community. These focused working groups allow leaders in those fields to concentrate their efforts and bring the most impactful contributions to the community. These working groups are open for anyone to attend and contribute.
The goal of this working group is to define how data is read into and written out from memory in MONAI. Such input and output requires consideration of (a) the research and application workflows in which MONAI will operate, (b) the importance of effectively utilizing all available data for deep learning development and evaluation, and (c) the critical significance of understanding and preserving the physical space that is represented by a medical image so that its clinical validity is preserved. The working group has defined three broad sets of requirements for input and output for MONAI: research I/O requirements, reproducible I/O requirements, and clinical I/O requirements.
Group Lead
Group Lead
A key part of Project MONAI is the ingestion of data. This working group will focus on defining support for bioinformatics (e.g., images), biomarkers, and metadata (e.g., patient demographics, outcomes). Part of the focus explores the use of ontologies, various representations of the data, and what is in-scope for MONAI and what is left to other projects to define. Some of the image representations, for example, include data relating to dimensionality, spacing, components/channels, data orientation, and time (multiple studies/prior exams).
This working group covers the topics related to data preprocessing and augmentation modules for MONAI. In terms of functionality, these include patient-level transforms, such as simulating body habitus variations, aging; spatial geometric transforms, intensity transforms and acquisition transforms such as simulating medical imaging modalities, artefacts. In terms of technical implementations, these include differentiability, multi-dimensional data representation, lazy evaluation, caching, reproducibility, etc
Group Lead
Group Lead
Federated learning is emerging as a promising approach for training AI models without requiring sites to share data. This is particularly important in medical use cases, since there are privacy and regulatory restrictions to sharing data. A number of different methods are being developed for federated learning, and they are currently silo efforts with limited synergy. The goal of this effort is to unify the disparate methods of federated learning in a common MONAI framework. This will be pursued through implemented federated learning use cases in MONAI as well as generation of best practices and documentation to enable the community to more broadly participate and implement common and interoperable approaches.
Every year, hundreds of new algorithms are published in the field of biomedical image analysis. While validation of new methods has long been based on private data, publically available data sets and international competitions (‘challenges’) meanwhile allow for benchmarking algorithms in a transparent and comparative manner. Recent research, however, has revealed several flaws related to common practice in validation. A core goal of this working group is, therefore, to provide the infrastructure and tools for quality-controlled validation and benchmarking of medical image analysis methods. In collaboration with the international biomedical image analysis challenges (BIAS) initiative, open technical challenges and research questions related to a variety of topics will be addressed, ranging from best practices to how to create incentives for sharing code and data to performance aspects and implementation efficiency.
Group Lead
Group Lead
Establish MONAI as a catalyst for scientific progress and real life impact through the following actions:
This working group focuses on outreach, to establish MONAI as a common software foundation that the Medical Imaging research and development community can build upon. This working group is charged with developing technical training and onboarding content, including tutorials, technical blogs, bootcamps, hackathons to educate and engage with researchers and developers. Another area of focus for the working group is to define the infrastructure and processes needed to support and grow a transparent, honest, and open community of MONAI contributors and adopters, including establishing best practices for giving appropriate recognition to those who are promoting and making significant contributions back to MONAI.
Group Lead
Group Lead
This working group aims to define how to close the existing gap from research and development to clinical production environments by bringing AI models into medical applications and clinical workflows with the end goal of helping improve patient care. The focus will include defining the open high-level functional architecture and determining which components and standard APIs are required. By collaborating with the MONAI developers, the group will move from requirements to implemented solutions.
Following a phased approach, the deploy working group will concentrate on what the end-to-end experience should look like first. Some of these challenges include:
Digital pathology as an imaging modality is a relatively new field compared to radiology imaging. Digital pathology focuses on Whole-Slide Images (WSIs), which are digital scans of tissue slides at the microscopic resolution, often at 250nm per pixel. WSIs are different from standard medical images due to their sheer size - a typical pathology image can be of the order of 150K x 100K pixels.
Recently, there have been significant advances in applying image analysis and machine learning to pathology images, also known as computational pathology. Despite those advances, there exists no standard pipeline for the preprocessing, analysis, and visualization of pathology images. This lack of standardization leads to a high barrier to entry and a lack of reproducibility for existing methods. This working group will aim to address those challenges around standardization and reproducibility.
Group Lead
Group Lead
The mission of the MONAI Stream working group is to advance MONAI as a platform for real-time AI research involving the multitude of interfaces, signals, analyses, simulations, and procedures associated with image-guided interventions and point-of-care applications. To fulfill that mission, we are investing in the integration of MONAI with robotic and surgical navigation systems such as ROS, DVRK, OpenIGTLink, and SlicerIGT, to meet the challenges associated with fusing AI with trackers, image acquisition systems, daVinci robots, AR/VR devices, and a multitude of other medical systems. Driving tasks include surgical event recognition, real-time 3D reconstruction from endoscopic video, collaborative camera control, and skill assessment. A major consideration is GPU resource management that supports multiple inference engines running in parallel as well as methods for streaming signals directly to GPU memory.
