Artificial intelligence is seen as revolutionizing how medical images are interpreted by helping medical professionals save time analyzing magnetic resonance imaging (MRI) scans, computerized axial tomography (CT or CAT) scans, and X-rays. However, the lack of accurate and reliable data to train neural networks for this purpose is a significant challenge facing deep learning scientists working in the medical community.

Now, for the first time, says Nvidia, researchers are using generative adversarial networks (GANs) to create synthetic abnormal brain MRIs that can be used to train neural networks.

“Data diversity is critical to success when training deep learning models,” say the researchers. “Medical imaging data sets are often imbalanced as pathologic findings are generally rare, which introduces significant challenges when training deep learning models. We propose a method to generate synthetic abnormal MRI images with brain tumors by training a generative adversarial network.”

The researchers used an Nvidia DGX system – which contains Nvidia Tesla V100 GPUs – with the cuDNN-accelerated PyTorch deep learning framework to train their generative adversarial network on data from two publicly available datasets of brain MRIs. One dataset contained thousands of 3D T1-weighted brain MRIs with Alzheimer’s disease, while the other contained about two hundred 4D brain MRIs with brain tumors.

GANs have previously been used in medical imaging, says the company, to generate a motion model from a single preoperative MRI, upsample a low-resolution image, create a synthetic head CT from a brain MRI, perform medical segmentation, and automatically align different types of MRIs, saving doctors hours.

“This offers an automatable, low-cost source of diverse data that can be used to supplement the training set,” say the researchers. “For example, we can alter a tumor’s size, change its location, or place a tumor in an otherwise healthy brain, to systematically have the image and the corresponding annotation.”

In addition, since the images are synthetically generated, there are no patient data or privacy concerns. Medical institutions can easily share data they generate with other institutions, say the researchers, creating millions of different combinations that can be used to accelerate the work.

For more, see “Medical Image Synthesis for Data Augmentation and Anonymization using Generative Adversarial Networks.” (PDF)


Related articles:
Nvidia supercomputing platform to ‘revolutionize’ medical imaging
Nvidia platform brings AI to next-gen medical instruments
Deep learning improves medical imaging
Google microscope combines AI, AR for real-time cancer detection
AI fixes grainy photos without needing to see clean ones

If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :    eeNews on Google News


Linked Articles