Artificial neuron network for autism study

Using a computational model that simulates the functioning of the cerebral cortex based on visual stimuli, the researchers determined the relationship between physiological and cognitive processes that may occur in autism spectrum disorders.

“A better understanding of the relationship between brain physiology and cognition or behavior is essential to designing new treatments, although there is still a long way to go to get to this point,” noted Rodrigo Echfest, PhD, a natural sciences doctor in Germany and a researcher from CONICET and Universidad Nacional del Littoral (UNL). ) at the Institute for Research in Computer Signals, Systems and Intelligence or “Sense (1)”, in the city of Santa Fe, Argentina.

A few years ago, Elizabeth Pellicano, of Macquarie University in Australia, and David Burr, of the University of Florence in Italy, suggested measuring or “balancing” sensory information from the outside world with prior information, and a way of representing possibilities about the outside world and about own expectations, occurring less effectively in Autistic people. For example, sensory perception is increased and expectations are dimmed.

Eschvist and her colleagues set out to try to understand why people with autism spectrum disorders had this different weight.

“To answer this question, we analyzed observations about the physiology of autism,” says the Santa Fe researcher, who has a PhD in the Computational and Biological Intelligence Laboratory (CBL) of the University of Cambridge, UK.

Arithmetic model

As a test bed, the study used a computational model developed by Echeveste with researchers from the University of Cambridge while they were staying at that institution.

“This model is a neural network trained using artificial intelligence techniques to process visual stimuli, which mimics the behavior of our primary visual cortex. When you measure how these artificial neurons behave while the network is “looking” at something, we see dynamics in the responses of the neurons that very similar to those in our cerebral cortex.”

An artistic reformulation of the concept of a computational model that simulates the functioning of brain cells. (Illustration: Jorge Monchi for NCYT de Amazings)

The optimal functioning of our brain requires a very delicate balance between excitatory neurons that, when activated, help turn other neurons on, and other inhibitory neurons that, when turned on, tend to turn off other neurons.

“Specifically, we wondered whether perceptual differences (between autistic and non-autistic people) in terms of information weight could be explained by differences in inhibition mechanisms, and for this, we used a computational model that simulated neurophysiological performance.” .

By weakening the inhibition in the artificial neural network, earlier expectations began to wane and the perception of stimuli increased. “So, at least in our model, these two views of autism would be two sides of the same coin,” the researcher said.

In recent years, there has been an explosion in neuroscience of using avant-garde techniques from the field of artificial intelligence to develop computational models that allow for a better understanding of the workings of the brain. Most models represent the normal functioning of the brain. Through this work, we show that this approach is also very useful for understanding sensory processing, for example, in people with autism,” Eschvist noted.

He added, “In a round trip with experimental neuroscience, these models, which are increasingly detailed and complex, allow hypotheses to be tested and generate predictions of the direction of future work, helping to focus experimental efforts, which are often difficult or expensive to implement.”

The title of the study is “Bridging Physiological and Cognitive Views of Autism by Bayesian Inference-Based Sampling”. It has been published in the academic journal Network Neuroscience. The authors also include Diego Mellon and Enzo Ferrante, also researchers from CONICET and UNL in Synchronization (i), and Ines Samingo, from CONICET at the Balseiro Institute and the Department of Medical Physics at the Bariloche Atomic Center. (Source: CyTA-Leloir Agency)