title:: Large scale AI in digital pathology without the heavy lifting | AWS Public Sector Blog
publisher:: Amazon Web Services (AWS)
people:: Moritz Widmaier, Martin Schulze, Florian Leiss, Ruben Cardenes, Razvan Ionasec
organization:: Ultivue, AWS, Slash-m GmbH
domain:: Healthcare, Artificial Intelligence, Digital Pathology, Medical Imaging, Cloud Computing
link:: https://aws.amazon.com/blogs/publicsector/large-scale-ai-digital-pathology-without-heavy-lifting/
Summary
This blog post discusses how the healthcare technology company Ultivue is leveraging AWS services like Amazon S3, AWS Storage Gateway, and Amazon SageMaker to enable large-scale AI-based analysis of high-resolution multiplex immunofluorescence tissue images for digital pathology applications. Ultivue combines multiplex tissue staining with computational pathology AI to deeply characterize immune cells, tumor cells, and their spatial relationships in tissue samples. Their solution handles massive multi-gigabyte, multi-channel image data on AWS while providing a simplified user interface through Amazon SageMaker Studio for medical experts without coding expertise. AWS enables scalable storage, computation, and user experiences to unlock new AI capabilities in digital pathology and precision medicine.
Data Points
- Ultivue provides multiplex immunofluorescence tissue assays and AI-based computational pathology
- Their multi-channel tissue images can reach 100GB with over 5 billion pixels
- Uses Amazon S3 for scalable storage, AWS Storage Gateway for file access
- Amazon SageMaker used for processing, training AI models, inference on images
- Amazon SageMaker Studio provides simplified UI for non-technical medical users
- Enables analysis of spatial biology data from multiplex imaging at large scale
- AI helps reveal immune-tumor interactions and predict therapy response
- Combines expertise in tissue assays and computational pathology AI
- Goal is to advance precision medicine through better tissue analysis
- Leverages AWS scalability while maintaining usability for domain experts