2023-05-05 02:20:57
By revealing the links between brain activity patterns and visual inputs, AI could eventually reveal ways to generate visual sensations in people with visual impairments.
Scientists at the Swiss Federal Institute of Technology in Lausanne have reconstructed, almost perfectly, a black and white video clip from brain activity data that was recorded, and then analyzed, with an artificial intelligence (AI) tool. This tool could eventually help visually impaired people.
Brain activity signals were collected while a group of 50 mice I watched nine times the 600 frames of a video. The 30-second clip showed a man running towards a car before opening the trunk. The recording of the electrical pulses of the neurons was carried out with metallic probes inserted in the cortical area involved in the processing of visual information. Some data on brain activity under a microscope was also collected.
Then, the researchers trained the AI named CEBRA to link this data to the clip. The team tested the ability of their trained AI to predict the order of frames within the movie. This time using brain activity data that was collected from the mice while they were watching the movie for the tenth time. This exercise revealed that the AI can predict the correct frame within a second 95% of the time.
The neuroscientists also trained their AI on brain data from individual mice. to test whether the reconstruction worked better with the individualized data. However, the AI predicted the frames of the movie that were now being watched with between 50% and 75% accuracy. The results of this work were published this Wednesday in Nature.
We recommend: Eternal youth? They manage to reverse aging in mice to be able to do it in humans
“Training the AI on data from multiple animals actually makes the predictions stronger, so you don’t need to train the AI on data from specific individuals to make it work for them,” commented Mackenzie W. Mathis, who led the study.
future applications
Mathis believes that by revealing the links between brain activity patterns and visual inputs, AI could eventually reveal ways to generate visual sensations in visually impaired people. “You can imagine a scenario where you really want to help someone who is visually impaired see the world in interesting ways by playing with the neural activity that would give them that sense of vision,” she stresses.
Shinji Nishimoto, neuroscientist at Osaka University, Japan, believes that this breakthrough could be a useful tool for understanding the neural codes that underlie our behavior. Nishimoto, who was not involved in this study, believes that it should be applicable to human data.
1683273771
#reconstruct #mice #world #artificial #intelligence