Scientists have developed a new artificial intelligence technology that can analyze brain activity and reconstruct images that a person is looking at, which could open the door to ways of understanding what a person sees without them describing it themselves.
During a study conducted by a team from the Weizmann Institute of Science, eight volunteers were asked to look at a series of pictures, including a baseball game, a dog looking out of a car window, and people walking across a snow-covered plain.
The scientists collected thousands of brain scans while viewing the images and used them to train a system they called "Brain-IT". The system learned how certain patterns of brain activity are associated with the colors, shapes, and objects that appear in the images.
After training, the system was able to analyze a new brain scan and reconstruct the image that the person was looking at, even when the image itself was not present in the training data, and with a remarkable degree of accuracy.
Professor Michal Irani, from the Weizmann Institute of Science, said that previous models were able to convert brain activity into images that preserved the overall meaning of the image, but they had difficulty reproducing basic details such as composition and colors.
She added that "Brain-IT" outperforms these models in reconstructing image content and details, and it only needs about one hour of brain scan data to learn to deal with a new person, compared to about 40 hours needed by other systems.
To test this, the scientists compared images produced by the system after training it on data collected over one hour with images produced after training it on data collected over 40 hours. The results were remarkably similar. The training also revealed 128 functional areas common to all people, which appear to have specific roles in processing visual information. Some of these areas are already known to neuroscientists, while others were discovered by the team for the first time.
Among the findings, scientists discovered a difference in the way images of places are processed within a region of the brain known as the PPA, where one part responds to internal scenes, while another part responds to external scenes.
The team doesn't plan to stop at images; they are currently working on developing the technology to decode audio information and are also exploring its potential use with video. Video presents a greater challenge due to the rapid change in image speed compared to the duration of functional magnetic resonance imaging (fMRI).
Irani said that overcoming these obstacles could make it possible in the future to use the technology to understand the content of dreams as well. Scientists are also working on developing systems based on electroencephalography (EEG), which measures the brain's electrical activity using sensors placed on the scalp.
The study's findings were presented at the "Computational Cognitive Neuroscience" conference, held in New York last month.
what is your opinion on this news ?do let us know in the comments on tha comment box
if you liked this news article please share on friends & family & don't forget to follow on website & social media