DOCOMO's New AI Generates Images of Predicted Changes in the Brain

Expected to support healthier lifestyles and improved brain health

NTT DOCOMO, announced today that it has developed an AI technology that predicts how the brains of adult people will physically change over time and, using a dataset of around 150,000 MRI brain scans, automatically generates images of how people's brains are expected to look in the future. The technology uses a machine-learning framework known as a generative adversarial network (GAN) to make predictions based on various factors.

The technology generates brain images by predicting changes in the volume of the hippocampus, which affects the brain overall as well as memory and dementia. It will be implemented in easy-to-use solutions that allow non-specialists to understand the state of brain health and ideally support healthy lifestyle habits and regular brain checkups.

The new technology was developed under the medical supervision of Dr. Yuhei Chiba, a psychiatrist, dementia specialist and founder of YUAD, and in collaboration with Biomy, Inc.

By 2025, about one in five elderly people in Japan is expected to be living with some form of dementia, underscoring the urgent need for an inclusive society in which dementia patients and their families can live with confidence and peace of mind. While it is possible for people to adopt healthy lifestyle habits and undergo regular brain check-ups, it can be difficult to imagine how lifestyle choices might affect one's own brain health.

DOCOMO is now developing simple ways to use lifestyle information collected through smartphones and surveys, aiming to encourage behavioral changes that promote brain health throughout society. DOCOMO envisions the integration of healthcare apps and brain health checkups (docks) to increase acceptance of healthier lifestyles and the need for brain docks as well as add value to brain-dock diagnoses, ultimately to support longer healthy lifespans and limit healthcare and caregiving costs.

This initiative is part of DOCOMO's Lifestyle Co-Creation Lab, a program through which DOCOMO is collaborating with various partners to evaluate life-enhancing technologies that could be incorporated into an “innovation co-creation platform,” which is expected to have useful applications in various industries. The technology will be exhibited at Docomo Open House '24 on January 17 and 18.

Technology that generates high-quality data by having a generator create data similar to real data and a discriminator model learning to distinguish between real and generated data through competitive training. YUAD is a registered trademark of Dr. Yuhei Chiba. “Annual Report of Aging Society,” The Cabinet Office, 2017.

About New AI that Generates Images of Predicted Changes in the Brain

The brain-image generation AI was developed with GAN, an image-generation AI model, to acquire and utilize a dataset of approximately 150,000 brain images from 500 individuals in the Open Access Series of Imaging Studies (OASIS), which provides datasets of brain neuroimages. The timing of when to input data, such as people's ages, during the model's training phase was carefully regulated to effectively reflect age-related details in the brain images generated by the technology.