NVIDIA AI Supercomputer Helps Generate World's Largest Brain Images Dataset
Researchers at King's College London have created a model and dataset of over 100,000 synthetic 3D brain images with the help of NVIDIA supercomputer Cambridge-1.
Created on May 31|Last edited on May 31
Comment
Many fields of research are lucky to have easy access to mountains of various types of data at their finger tips, easily accessible for use in data science and machine learning. However, the medical sciences are often not as fortunate, as accumulating data can be an expensive process, contain considerable locality bias, and be subject to patient privacy concerns.
One researcher, Jorge Cardoso, along with his team, set out to fill in one of these data gaps. The London-based researchers developed an AI model to generate a dataset of 100,000 synthetic brain images, all freely available for interested medical researchers. Though they are "synthetic", meaning they're not real brain scans of a human being, they still exhibit the key biological characteristics that real brains have, meaning they're perfectly suitable for use in research.
The tools for the job
To generate their model and dataset, they employed the power of NVIDIA's Cambridge-1 supercomputer, a massive processing structure dedicated to AI research in healthcare featuring 640 NVIDIA A100 Tensor Core GPUs.
The powerhouse supercomputer was driven by MONAI, a framework for AI computing and medical imaging, alongside built-in NVIDIA technologies.
By their powers combined, the dataset of 100,000 synthetic 3D brain-scan images (featuring 16 million 3D pixels each) was able to be created. The dataset and models will be hosted by Health Data Research UK.
Find out more
Add a comment
Tags: ML News
Iterate on AI agents and models faster. Try Weights & Biases today.