The blurry line between vision and imagination uncovered using AI

Articles

Mental imagery occurs when we can represent perceptual information in our minds without the required sensory input. The small neural differences between vision and mental imagery have been unclear and troubling for neuroscientists.

However, using artificial intelligence (AI), researchers have exposed the neural pathway that creates vision and mental imagery in both the human brain and in a computer network, emphasising the power of computational neuroscience today.

Researchers from the Medical University of South Carolina used machine learning to analyse vision and mental imagery. This is a method where machines are designed to automatically learn and improve from experience, allowing them to learn for themselves. The computer network used is similar to a biological network, where the different sections of the network represent groups of neurons in the brain, each with a different function.

The team trained the computer to view images and then had the computer imagine images. 3 human participants were then asked to follow the same steps: view images on a screen and then imagine them, whilst analysed using brain imaging (fMRI).

Repeating the process allowed the team to identify the sections of the human brain that were active and contrast these with the sections from the computer model.

The results from the human brain demonstrated that during vision and mental imagery the pathway from the retina of the eye to the visual cortex are activated and extend further. Whereas during mental imagery only, that same pathway is diffuse. The computer areas that were activated were the same as in the human brain and emphasised that the differences between vision and mental imagery resides between the retina and the visual cortex.

diagram of brain

Source : Principles of Psychology, Pressbooks (2019). Queen’s University.  https://ecampusontario.pressbooks.pub/testbookje/chapter/seeing/

These results prove that an encoding model for mental images can predict human brain activity during vision and mental imagery, demonstrating the importance of computational neuroscience in today’s research and the direction in which it can be taken when investigating mental imagery.

The ‘fuzzy’ and diffuse pathway found during mental imagery allows individuals to distinguish between states, for example, when they are awake or when they are or were dreaming. However, in certain psychiatric or sleep disorders this distinction is not as obvious. Therefore, these findings could have large implications in the study of psychiatric disorders that impact mental imagery, such as post-traumatic stress disorder (PTSD), obsessive-compulsive disorder (OCD) and schizophrenia, where frequent intrusive mental imagery can be debilitating. As well as in the study of sleep disorders, such as narcolepsy, where dream-reality confusion is so frequent.

Original paper: J. L. Breedlove, G. St-Yves, C. A. Olman and T. Naselaris (June 2020). Current Biology Vol. 30, Issue 22.
https://www.sciencedirect.com/science/article/pii/S0960982220304942

Artwork by Garth Laidlaw entitled ‘Child’s Imagination’.
https://www.inprnt.com/gallery/garthlaidlaw/

Leave a Reply

Your email address will not be published. Required fields are marked *