Scientists at Israel’s Weizmann Institute of Science have developed an artificial intelligence model called Brain-IT that is designed to reconstruct visual scenes based solely on brain activity.
According to its developers, Brain-IT identifies patterns of brain activity shared by most people and uses them to translate mental visual scenes into images. The researchers say the technology could eventually help restore communication for people who have experienced severe paralysis.
Developing the model required researchers to work with limited data, including large collections of images and functional magnetic resonance imaging scans, or fMRI, recorded while participants viewed different scenes.
The team developed what it describes as a two-way “bilingual dictionary” between images and brain activity. Rather than creating only a decoder that reconstructs an image from an fMRI scan, researchers also developed an encoder that predicts brain activity from an image.
The encoder identified 128 functional regions shared among participants, each involved in specific aspects of visual processing.
More Read
According to research director Michal Irani, the encoder could eventually help scientists better understand how different regions of the brain work together and how visual perception takes shape.
