From NVIDIA: 2024-09-20 10:00:49

A committee of experts is using NVIDIA-powered federated learning to train AI models for tumor segmentation in medical imaging. Federated learning allows collaboration without sharing sensitive data. The group from SIIM is working on a project involving renal cell carcinoma data from six medical centers, optimizing models for accuracy and speed.

In a federated learning framework, global model parameters are exchanged instead of data, improving privacy and security. The team is experimenting with model architectures and hyperparameters to optimize training for kidney cancer imaging studies. They are also exploring AI-assisted annotation with NVIDIA MONAI to improve annotation accuracy and efficiency.

The team plans to publish their methodology and pretrained model after completing the project. They are using MONAI Label for image labeling, hosted on the Flywheel platform. The NVIDIA Academic Grant Program supports their research with resources like RTX A5000 GPUs. Future projects will focus on data science, graphics, and federated learning applications.



Read more at NVIDIA: Medical Centers Tap AI, Federated Learning for Better Cancer Detection