Machine-Learning Algorithm Uses Brain Signals To Predict What Mice See
May 05, 2023 By Awanish Kumar
(Image Credit Google)
A mouse watches a black-and-white movie (Image credit : EPFL)
We are a long way from creating the technology that will allow us to see what other people view from their eyes, despite a team of researchers previously being able to
interpret thoughts using artificial intelligence and MRI scans.
However, a group of Swiss researchers has now taken a significant step towards making this a reality.
When a mouse was made to watch a black-and-white movie as part of a demonstration, scientists from the
Swiss Federal Institute of Technology (EPFL) in Lausanne used a new AI tool to reconstruct what the mouse had seen.
The cutting-edge machine learning technique known as CEBRA can anticipate complicated information, such as what mice observe, by revealing the hidden structure in data recorded from the brain.
“This work is just one step towards the theoretically-backed algorithms that are needed in neurotechnology to enable high-performance BMIs (brain-machine-interfaces),” says Mackenzie Mathis, EPFL’s Bertarelli Chair of Integrative Neuroscience, who spearheaded the
study.
For the investigation, Mathis and her group used an animal model. The 50 mice were given a 30-second movie clip to watch while the researchers recorded the brain activity. Nine times were required for the mice to see the film. The CEBRA artificial intelligence (AI) program, developed by the researchers, was then trained to connect the brain data to the video.
CEBRA gains the ability to associate specific frames with brain activity throughout the training phase. According to the
press release, CEBRA operates well with less than 1% of the visual cortex's neurons, despite the fact that mice have a visual cortex with about 0.5 million neurons.
[caption id="attachment_169548" align="aligncenter" width="1155"]
For learning the latent (i.e., hidden) structure in the visual system of mice, CEBRA can predict unseen movie frames directly from brain signals alone after an initial training period mapping brain signals and movie features. Credit: Neuroscience News[/caption]
After watching the video ten times, the team put CEBRA to the test to see if it could accurately anticipate the clip's frame order based on information about brain activity. They were attempting to determine whether they could decode the mice's natural footage on a frame-by-frame basis. They succeeded in deciphering with higher than 95% accuracy.
“This algorithm is not limited to neuroscience research, as it can be applied to many datasets involving time or joint information, including animal behavior and gene-expression data. Thus, the potential clinical applications are exciting.”
By Awanish Kumar
I keep abreast of the latest technological developments to bring you unfiltered information about gadgets.