NVIDIA Clara Imaging
Today, GPUs are found in almost all imaging modalities, including CT, MRI, X-ray, and Ultrasound - bringing compute capabilities from IT data centers and clouds to the edge devices. With the boom of deep learning research in medical imaging, more efficient and improved approaches are being developed to enable AI-assisted workflows.
To develop these AI-capable applications, the data needs to be made AI-ready. NVIDIA Clara’s AI-Assisted Annotation does so by providing APIs and a toolkit to bring AI-assisted annotation capabilities to any medical viewer. Post annotation, data scientists and researchers need to build a robust AI model. To enable this, NVIDIA Clara Train includes techniques like AutoML, privacy-preserving federated learning, and Transfer Learning. After training, you need to bring your model to the right clinical environments, and you can do this with NVIDIA Clara Holoscan. Clara Holoscan reduces engineering investment for building high-performance AI Inference Platforms.
Medical Imaging AI Workflow
NVIDIA Clara for Medical Imaging helps accelerate a typical imaging workflow that starts with labeling data, creating AI models, developing applications that include one or several AI models, and finalizing their deployment at an institution or across multiple institutions.
NVIDIA AI-Assisted Annotation (AIAA) with DeepGrow allows fast labeling of data with a few clicks on an area of interest and in a few slices for CT and MRI. AIAA helps speed up image labeling, which can be a time-consuming and tedious task prone to user variation.
NVIDIA Clara Train SDK provides over 20 pre-trained AI models for radiology and digital pathology that are ready to download and use immediately or be tailored to an institution’s data via the transfer learning toolkit or trained on various institutions’ data via our Federated Learning framework. With Federated Learning, a trained local model can be trained on diverse patient data across regions and institutions by sharing a subset of model weights without sharing patient information.
NVIDIA Clara Holoscan is a hybrid computing platform for medical devices that combines hardware systems, optimized libraries, SDKs, and core microservices needed to develop and run end-to-end streaming and imaging applications, from embedded to edge to cloud.
AI-Assisted Image Labeling
Clara Train AI-Assisted Annotation (AIAA)
NVIDIA AI-Assisted Annotation (AIAA) provides Client APIs and Annotation server that can seamlessly enable any medical viewer to be AI-powered and can speed up the creation of high-quality annotated datasets that can be used to train robust AI algorithms.
- Increases the efficiency of annotation workflow by 10x
- Provides pre-trained models & intelligent algorithms to kick start annotation with a few clicks
- Configurable allowing users to define custom inference logic
- Provides robust server-client architecture for developers to scale and extend their viewer applications with AI-powered annotation workflow
AI Model Training
Clara Train SDK
NVIDIA Clara Train SDK provides a training framework to help accelerate deep learning training and inference for medical imaging use cases. It allows medical imaging researchers and developers to implement new models using a high-level intuitive API quickly.
Clara Train is built on MONAI Core, a PyTorch-based framework for deep learning in healthcare imaging, and abstracts these domain-specialized build blocks from MONAI into MMAR (Medical Model Archive) format.
- Provides state of the art pre-trained models to help you quickly get started
- Increases the training performance by up to 50x with domain-specific GPU optimization for 3D Imaging model
- Provides powerful techniques like AutoML, that enable automatic parameter tuning and make iterative experimentation faster.
- Clara Train’s Federated Learning is powered by Nvidia's stanalone python library nvFlare that provides an easy way of provisioning distributed clients for multi-institutional collaboration for medical imaging use-case.
Clara Holoscan SDK
NVIDIA Clara Holoscan SDK provides healthcare-specific acceleration libraries, pre-trained AI Models, and Reference Applications for computational medical devices in ultrasound, endoscopy, surgical robotics, digital pathology, radiation therapy, patient monitoring CT, MRI, X-ray, Genomics, Microscopy.
Clara Holoscan for traditional medical imaging inference use cases, including diagnosis, quality assurance and research validation, is still in development, you can learn more on the Holoscan SDK Page.
If you are looking to get started today, we encourage you to check out MONAI Deploy. It’s an open-source, easy-to-get-started set of SDKs, reference applications and microservices, created by the Project MONAI community.
MONAI Deploy aims to become the de-facto standard for developing, packaging, testing, deploying, and running medical AI applications in clinical production. MONAI Deploy is focused on defining the journey from research innovation to clinical production environments in hospitals. The guiding principles are:
- Create tangible assets: tools, applications, and demos/prototypes.
- Radiology first, then other modalities like Pathology.
- Interoperability with clinical systems. Starting with DICOM, then FHIR.
Developer Blogs on NVIDIA Clara
Dive into the features and capabilities of the Clara Imaging application framework.
Open Source Framework for Medical Imaging
MONAI, an open source framework for healthcare builds on best practices from existing tools like NVIDIA Clara, NiftyNet, DLTK, and DeepNeuro. In this webinar, learn how you can engage and contribute to this framework.
Clara Imaging SDKs for COVID-19 Research
Learn the latest capabilities and reference workflows that developers can use to accelerate their research and development.
Hands-On Clara in the Medical Imaging Ecosystem
In this webinar, watch a real-time demonstration of a medical imaging workflow using open source components and the Clara Deploy application framework.
NVIDIA Clara Developer Sessions at GTC
Learn more about the Clara Train and Clara Deploy applicaiton frameworks in these deep dive on-demand technical sessions. These sessions focus on federated learning, AI model training, scalable and modular deployment of AI models, and connecting it all to a medical imaging ecosystem.
Get Started with NVIDIA Clara
NVIDIA's David Nola walks through how to integrate Clara Train and Clara Deploy medical imaging tools into existing AI infrastructure.
How Clara Federated Learning Works
Learn how NVIDIA Clara Federated Learning enables institutions to collaboratively build robust AI models for medical imaging while keeping patient data private.
An Overview of NVIDIA Clara Deploy
Take a closer look at how NVIDIA Clara Deploy works with this demo using a multi-AI imaging pipeline.
Deep Learning Training for Medical Imaging
Get hands-on training in AI for healthcare through the NVIDIA Deep Learning Institute (DLI). Take online courses like Medical Image Classification Using the MedNIST Dataset to get an introduction to deep learning for radiology or learn how to use generative adversarial networks (GANs) in Data Augmentation and Segmentation with Generative Adversarial Networks for Medical Imaging.
View All Courses