AI system creates synthetic brain scans

Newly accessible brain scans can be used to help train neural networks and further develop advances in medical research using AI.

Artificial intelligence has had a huge impact on how medical images are understood. This helps medical professionals save time analysing magnetic resonance imaging, CT scans and X-rays. Despite this technological assistance, a significant challenge for scientists is still the lack of accessible, accurate and reliable data to train their neural networks. Thanks to researchers at an American technology company NVIDIA, scientists are now using Generative Adversarial Networks (GANs) to generate abnormal brain MRIs that can be used to train neural networks.

Researchers from NVIDIA’s Mayo Clinic, and the MGH & BWH Center for Clinical Data Science have developed a deep learning-based model that can generate accurate synthetic images that can be used for training an artificial intelligence system. Additionally the NVIDIA DGX-system allows team members to train their generative adversarial network on data. One dataset contains thousands of 3D T1-weighted brain MRIs with Alzheimer’s. The other scans contain around 200 4D brain MRIs with brain tumours.

The team based their image-to-image translation method on the pix2pix model, previously developed by NVIDIA researchers. As the images aren’t real, there is no patient data or privacy concerns. This enables medical institutions to share data they generate with other institutions. In turn, creating millions of different combinations that can be used to accelerate the work. NVIDIA hopes the model can help scientists generate new data that can be used to detect abnormalities and therefore save lives.

AI has been used in a multitude of methods to help those who have health and medical needs. An AI-enabled tool has been created to help predict the likelihood of IVF success for couples looking to conceive, for example. Elsewhere, an AI mini robot helps reduce anxiety in children with autism. How could AI-enabled tools impact your day-to-day life both personally and professionally?