Scientists at the Swiss Federal Institute of Technology (EPFL) have developed a machine learning algorithm called CEBRA that can decode and reconstruct what mice see by analyzing their brain signals.
Scientists grow antlers on mice, hope to regrow human limbs
A group of Chinese scientists has transplanted deer genes onto a mouse, causing it to grow antlers. Deer shed and regrow their antlers annually,…
The groundbreaking algorithm could help reveal hidden structures within data recorded from the brain, with potential applications in visual neuroprosthetics and paralysis treatments.
EPFL neuroscientist Mackenzie Mathis explained that CEBRA can build a latent representation of the data and predict the sequence of frames a mouse is watching with over 95 percent accuracy.
This is considered a significant step toward decoding brain signals for brain-machine interface applications.
The researchers utilized open-access data from the Allen Institute in Seattle, Washington, for video decoding, and brain signals were obtained either directly from the visual cortex or using genetically modified mice with glowing green neurons.
Though CEBRA cannot yet fully reconstruct what humans see based on brain signals alone, it is a significant advancement in that direction, with potential clinical applications beyond neuroscience.