Advancing cancer detection with AI and federated learning: A committee of experts from leading U.S. medical centers and research institutes is leveraging NVIDIA-powered federated learning to enhance AI models for tumor segmentation.
- The project aims to evaluate the impact of federated learning and AI-assisted annotation on training AI models for more accurate cancer detection.
- Federated learning allows organizations to collaborate on AI model development without compromising data security or privacy, as sensitive data remains on local servers.
- The technique is particularly valuable in medical imaging, where privacy constraints and rapid AI development make traditional data-sharing methods increasingly challenging.
Key participants and technology: The Society for Imaging Informatics and Medicine (SIIM) Machine Learning Tools and Research Subcommittee is spearheading this initiative, utilizing NVIDIA’s technologies and support.
- Participating institutions include Case Western, Georgetown University, Mayo Clinic, UC San Diego, University of Florida, and Vanderbilt University.
- NVIDIA FLARE (NVFlare), an open-source framework, is being used for its robust security features and advanced privacy protection techniques.
- The NVIDIA Academic Grant Program provided four NVIDIA RTX A5000 GPUs to support the project, distributed across participating research institutes.
Project focus and methodology: The study concentrates on renal cell carcinoma, a type of kidney cancer, using medical imaging data from multiple centers.
- Six medical centers each contributed approximately 50 medical imaging studies for the project.
- The federated learning approach involves exchanging model parameters rather than raw data, preserving privacy and security.
- The team is experimenting with various model architectures and hyperparameters to optimize training speed, accuracy, and efficiency.
AI-assisted annotation phase: To further improve the model’s performance, the team is incorporating AI-assisted annotation using NVIDIA MONAI.
- This phase aims to evaluate how model performance differs when trained on AI-assisted annotations compared to traditional manual labeling methods.
- MONAI Label, an image-labeling tool, is being used to develop custom AI annotation apps, potentially reducing the time and effort required for dataset creation.
- Experts will validate and refine the AI-generated segmentations before using them for model training.
Data management and future plans: The project utilizes advanced data management solutions and aims to contribute to the broader medical AI community.
- Data for both manual and AI-assisted annotation phases is hosted on Flywheel, a medical imaging data and AI platform with integrated NVIDIA MONAI capabilities.
- Upon completion, the team plans to publish their methodology, annotated datasets, and pretrained model to support future research in the field.
Broader implications for medical AI: This collaborative effort highlights the growing importance of federated learning and AI-assisted tools in advancing medical imaging and cancer detection.
- The project demonstrates how federated learning can overcome data sharing challenges while maintaining compliance with privacy regulations like GDPR and HIPAA.
- By combining federated learning with AI-assisted annotation, the team aims to address issues of data uniformity across different medical centers, potentially improving overall annotation accuracy and model performance.
- The open publication of results and resources could accelerate the adoption of these advanced AI techniques throughout the medical field, potentially leading to more accurate and efficient cancer detection methods in the future.
Medical Centers Tap AI, Federated Learning for Better Cancer Detection
