NYU Langone neuroscientists identified the brain region likely responsible for recognizing images after seeing them once, even if the visuals are blurred or distorted. The ability, also called “one-shot perceptual learning,” stems from the high-level visual cortex — laying the groundwork for future research on treating neurological disorders.

To study this, researchers showed participants “Mooney images” — low-quality, grayscale pictures of objects — before showing them a clear version of the same object. Once a person sees the original image, they become twice as good at recognizing the Mooney images on a second look.

Jonathan Shor, a first author of the February study and doctoral student at the NYU Langone Institute for Translational Neuroscience, told WSN that this ability does not work like traditional memory. In a previous experiment, subjects lacking parts of their medial temporal lobe — a brain region necessary for making new memories — could successfully identify a Mooney image a week after seeing the clear version, indicating a separate system is responsible for this storage of images. 

“If you see a very corrupted image of a certain object, you might not be able to recognize it,” said Tandon doctoral student Xujin Chris Liu, another first author on the paper. “Later, you might see some kind of hint — like an ‘aha’ moment — that causes you to see it and this knowledge gets stored in your brain for a long time, very robustly. But we don’t yet know why and where this is stored in the brain.”

Identifying where this storage occurs could help treat hallucinations — common symptoms of neurological disorders like schizophrenia and Parkinson’s disease — which arise from an overactivation of the visual system. Additionally, better understanding the nature of this brain region may further research into the integration of “one-shot perceptual learning” into machine learning models of human perception.

Researchers first altered how the original image was presented — including rotations, position and size — to see its impact on the patients’ ability to correctly identify a Mooney image on second reference. They found that resizing the Mooney image had no effect, but all other manipulations worsened the participants’ ability to pinpoint the correct subject. This suggested activity within the HLVC, which is responsible for processing larger areas of visual space.

The second part of the study used electrodes to record electrical signals from different parts of the visual cortex. By monitoring both the activity and the sequence of signaling, researchers pointed to the HLVC region as the source of these flashes.

Due to the procedure’s invasive nature, Shor said this step in the research process could only be done with consenting individuals already being treated for seizures with implanted electrodes, which is why they were only able to test 19 participants across six years. 

“Unlike a typical study where you can put some flyers out or try and recruit people from the general population, I’m waiting for patients who the doctors deem appropriate to even have this procedure done,” Shor said.

In the final step, researchers built an artificial intelligence model of the visual system in an attempt to map out how our brains might be wired. After being trained on Mooney images, the model mirrored how humans perceive one-shot visuals. This allowed them to compare the steps of the model to corresponding regions of the brain, pointing to the HLVC as the site of signaling. 

“When your brain learns something, when you have an ‘aha’ moment, you don’t just temporarily change how you represent images, you have synaptic changes that change how your brain connects — our model does not do that,” Liu said, referring to limitations within the AI model.

Despite the findings pointing to the HLVC as the source of storage, Shor said a definitive answer would come from a future experiment that interferes with the function of the region.

“Some sort of perturbation to the neural processing in that area is something that my lab has thought about,” Shor said. “The natural experiment would be to apply it to this area that we identified and see if people no longer can recognize the degraded images, even though maybe they just saw the original unambiguous image. That would be a good next step.”

Contact Grayson Stotz at [email protected].