Weizmann AI Model Reconstructs Viewed Images From Brain Scans
Brain-IT needs only 1 hour of fMRI data from a new subject to match results other methods achieve with 40 hours of recording.
Researchers at the Weizmann Institute of Science have developed an AI model called Brain-IT that can reconstruct images a person is viewing from fMRI brain scans, according to MIT Technology Review. The model can also predict what an fMRI brain scan would look like for a specific image.
Brain-IT requires only 1 hour of fMRI data from a new subject to match results other methods achieve with 40 hours of recording. The team trained the model on thousands of brain scans from the Natural Scenes Dataset, which came from eight volunteers.
The researchers identified 128 functional regions in the brain that activate for specific visual content. The work is documented on GitHub and in a paper presented at a scientific conference earlier this year.
What the team says about 'mind-reading'
Michal Irani said in an interview with MIT Technology Review that "mind-reading" is a "cute, jazzy name" and that the technology does not capture thoughts, memory or language.
Irani said the team now wants to see if similar results can be achieved with auditory information and, in the future, dreams. The researchers are also looking into whether a simpler EEG device could deliver similar results, according to MIT Technology Review.
Quick answers
What is Brain-IT?
Brain-IT is an AI model developed by researchers at the Weizmann Institute of Science that reconstructs images a person is viewing from fMRI brain scans and can predict what an fMRI brain scan would look like for a specific image.
How much fMRI data does Brain-IT need?
It requires only 1 hour of fMRI data from a new subject to match results other methods achieve with 40 hours of recording.
What was Brain-IT trained on?
The model was trained on thousands of brain scans from the Natural Scenes Dataset, which came from eight volunteers.