Introduction to NVIDIA MONAI Cloud APIs

The NVIDIA MONAI Cloud APIs offer a comprehensive suite of tools designed to enhance workflows in medical imaging through interactive annotation, training, and inference. Here's an overview of their key features and how to get started:

What are NVIDIA MONAI Cloud APIs?

NVIDIA MONAI Cloud APIs is an NVIDIA-managed solution, focusing on making the medical imaging lifecycle easier with service integration from annotation, to training, to inference. With a foundation in the MONAI ecosystem, these APIs offer a seamless interface to researchers and data scientists aiming to deploy, manage, and refine AI models in medical imaging applications.

Key Features

1. Interactive AI Annotation:

   • Utilizes the VISTA-3D foundation model.
   • Supports interactive AI annotation, enabling ongoing improvement of AI models through user feedback and new data integration.

2. Auto3DSeg:

   • Focuses on automating the process of 3D segmentation model creation.
   • Implements best practices to deliver state-of-the-art segmentation performance

3. VISTA-3D Model:

   • A robust foundation model, trained on 132 classes encompassing various organs and tumors.
   • The model's training and validation involved CT images from over 4,000 patients, ensuring its high accuracy and reliability for medical image analysis.

4. VISTA-2D Model:

   • VISTA-2D is a cutting-edge model specifically designed for cell image segmentation.
   • Built on the flexible Meta SAM architecture and adapted for cell imaging, VISTA-2D outperforms the current leading model, Cellpose 2.0. It can be fine-tuned on customer-specific data to deliver unparalleled accuracy in segmenting cells and employs a more generalized approach that adapts more easily to different cell types.

5. MAISI Model

   • The Synthetic Data Generation Model is an advanced generative AI model that creates high-quality 3D CT images with or without anatomical annotations.
   • It generates high-resolution images (512 × 512 × 512) with a voxel size of 1.0 × 1.0 × 1.0 mm³ and supports up to 132 anatomical classes, including organs and tumors.
   • This model excels in data augmentation, creating realistic medical imaging data to supplement limited datasets due to privacy concerns or rare conditions.

How to Get Started

To begin using the NVIDIA MONAI Cloud APIs, you can sign up for early access by filling out this survey: NVIDIA MONAI Cloud API Early Access.