—
In the realm of modern healthcare, the importance of medical imaging cannot be overstated. Each year, approximately 3.6 billion medical imaging procedures, such as X-rays, CT scans, MRIs, and ultrasounds, are performed globally to diagnose, monitor, and treat a wide array of medical conditions. These tests are critical tools for healthcare professionals, but the sheer volume presents a challenge—how to efficiently process and evaluate these images to help doctors manage their workloads and improve patient outcomes.
Enter NVIDIA’s MONAI, an open-source platform aimed at revolutionizing the world of medical imaging through the power of artificial intelligence (AI). MONAI stands as a bridge between the realms of healthcare and data science, enabling the creation of deep learning models and deployable applications tailored for medical AI workflows. By facilitating collaboration between doctors and data scientists, MONAI is unlocking the potential of medical data to enhance diagnostic and treatment processes.
A significant development in this field was announced at the annual meeting of the Radiological Society of North America (RSNA). NVIDIA revealed that Siemens Healthineers, a global leader in medical technology, has adopted MONAI Deploy. This module within MONAI is designed to streamline the transition from research to clinical application, enhancing the speed and efficiency with which AI workflows are integrated into medical imaging systems.
Siemens Healthineers has already made a substantial impact with its enterprise imaging platforms, Syngo Carbon and syngo.via, which have been installed in over 15,000 medical devices worldwide. These platforms assist clinicians in analyzing and deriving insights from diverse medical images. However, the process of deploying AI applications in clinical settings can be complex due to the variety of frameworks developers use when building these applications.
MONAI Deploy simplifies this process. With just a few lines of code, it allows for the creation of AI applications that can operate in any environment. This tool is a comprehensive solution for developing, packaging, testing, deploying, and running medical AI applications in clinical settings. It significantly streamlines the integration of medical imaging AI applications into clinical workflows.
The integration of MONAI Deploy into the Siemens Healthineers platform has drastically reduced the time required to bring trained AI models into clinical settings. What once took months can now be accomplished in just a few clicks. This rapid deployment capability enables researchers, entrepreneurs, and startups to deliver their applications to radiologists more swiftly, ultimately benefiting patients.
Axel Heitland, head of digital technologies and research at Siemens Healthineers, emphasized the advantages of this accelerated AI model deployment. "By speeding up AI model deployment, we empower healthcare institutions to leverage the latest advancements in AI-based medical imaging faster than ever," he stated. "With MONAI Deploy, researchers can quickly tailor AI models and transition innovations from the lab to clinical practice, providing thousands of clinical researchers worldwide access to AI-driven advancements directly on their syngo.via and Syngo Carbon imaging platforms."
The integration of MONAI-developed applications into these platforms can further streamline AI integration. These applications are readily available on the Siemens Healthineers Digital Marketplace, allowing users to browse, select, and seamlessly incorporate them into their clinical workflows.
MONAI Ecosystem Boosts Innovation and Adoption
Celebrating its five-year anniversary, MONAI has achieved remarkable milestones, including over 3.5 million downloads, contributions from 220 individuals worldwide, recognition in more than 3,000 publications, 17 MICCAI challenge wins, and usage in numerous clinical products.
The latest release, MONAI v1.4, brings updates that offer researchers and clinicians even more opportunities to leverage MONAI’s innovations and contribute to Siemens Healthineers’ Syngo Carbon, syngo.via, and the Siemens Healthineers Digital Marketplace.
Among the enhancements in MONAI v1.4 and related NVIDIA products are new foundation models for medical imaging. These models can be customized in MONAI and deployed as NVIDIA NIM microservices. The following models are now generally available as NIM microservices:
- MAISI (Medical AI for Synthetic Imaging): A generative AI foundation model capable of simulating high-resolution, full-format 3D CT images and their anatomical segmentations.
- VISTA-3D: A foundation model for CT image segmentation, offering accurate performance for over 120 major organ classes. It also features effective adaptation and zero-shot capabilities to learn to segment novel structures.
In addition to these major features in MONAI 1.4, the new MONAI Multi-Modal Model (M3) is now accessible through MONAI’s VLM GitHub repository. M3 is a framework designed to extend any multimodal large language model (LLM) with medical AI expertise, incorporating trained AI models from MONAI’s Model Zoo. The VILA-M3 foundation model, now available on Hugging Face, demonstrates the power of this framework by offering state-of-the-art radiological image copilot performance.
MONAI Bridges Hospitals, Healthcare Startups, and Research Institutions
MONAI’s impact extends across leading healthcare institutions, academic medical centers, startups, and software providers worldwide. Notable adopters and contributors include:
- German Cancer Research Center: Leads MONAI’s benchmark and metrics working group, providing metrics for measuring AI performance and guidelines for their usage.
- Nadeem Lab at Memorial Sloan Kettering Cancer Center (MSK): Pioneered cloud-based deployment of AI-assisted annotation pipelines and inference modules for pathology data using MONAI.
- University of Colorado School of Medicine: Developed MONAI-based ophthalmology tools for detecting retinal diseases using various imaging modalities. The university also plays a key role in federated learning developments and clinical demonstrations using MONAI.
- MathWorks: Integrated MONAI Label with its Medical Imaging Toolbox, bringing AI and AI-assisted annotation capabilities to thousands of MATLAB users in academia and industry.
- GSK: Exploring MONAI foundation models, such as VISTA-3D and VISTA-2D, for image segmentation.
- Flywheel: Offers a platform that includes MONAI for streamlining imaging data management, automating research workflows, and enabling AI development and analysis, catering to research institutions and life sciences organizations.
- Alara Imaging: Published work on integrating MONAI foundation models, such as VISTA-3D, with LLMs like Llama 3 at the 2024 Society for Imaging Informatics in Medicine conference.
- RadImageNet: Exploring MONAI’s M3 framework to develop cutting-edge vision language models that utilize expert image AI models from MONAI to generate high-quality radiological reports.
- Kitware: Provides professional software development services around MONAI, helping integrate it into custom workflows for device manufacturers and regulatory-approved products.
Researchers and companies are also utilizing MONAI on cloud service providers to run and deploy scalable AI applications. Cloud platforms offering access to MONAI include AWS HealthImaging, Google Cloud, Precision Imaging Network (part of Microsoft Cloud for Healthcare), and Oracle Cloud Infrastructure.
For more detailed information, visit the official Siemens Healthineers pages on syngo.via, Syngo Carbon, and the Digital Marketplace.
For more Information, Refer to this article.
