A group of scientists from the Federal Polytechnic School of Lausanne (Switzerland) has developed a system based on artificial intelligence algorithms that can interpret rodent brain signals in real time and reconstruct what the mouse sees from them. In particular, the scientists were able to reconstruct the video shown by the mice using brain signals.
Image source: youtube.com/@epfl
The scientists called their AI algorithm CEBRA, and it was trained by comparing neural activity and certain video frames. This allowed him to subsequently predict and reconstruct the video clips the mouse is watching. In the demo video, an example was a fragment of a black-and-white film shot in the sixties, in which a man runs up to a car and opens its trunk. The other screen shows the video sequence restored by CEBRA – the recordings are almost identical, although the image twitches periodically on the second one.
Registration and measurement of the brain activity of the rodent was carried out using electrodes connected to the region of the visual cortex of their brain; as well as with the help of optical probes of genetically modified individuals, whose neurons were highlighted in green when information was transmitted. The mice were given movies to watch and their brain activity was recorded in real time, comparing the two data streams and thereby teaching CEBRA – the algorithm learned which brain signals were associated with specific frames of the tape.
The AI algorithm then received as input an unfamiliar stream of brain activity from a mouse that was watching a different piece of video. Based on this, the CEBRA system itself was able to restore the frames corresponding to these signals in real time, which the scientists combined into a separate film.
If you notice an error, select it with the mouse and press CTRL + ENTER.