We better understand how autistic brains process faces differently, thanks to artificial intelligence

New research could help us understand how the brains of autistic people have a harder time recognizing emotions in facial expressions.

Image from Pixabay.

Facial expressions are one of the main ways people convey their emotions to those around them. A smile is a good indicator of happiness; eye rolls are a pretty reliable sign that someone is getting frustrated. However, autistic people can have a hard time actually picking up on these displays.

We don’t know exactly why this is. New research focused on artificial intelligence could finally help us figure out why.

Inner workings of the brain

As far as we know, there are two brain regions that can explain where the differences in processing lie between the typical and autistic brain. One is the inferior temporal (IT) cortex, which handles facial recognition. Another is the amygdala, which takes in information from the IT cortex and interprets the emotional content of the expressions it perceives.

To understand the extent to which these two areas are involved in the differences in processing, Kohitij Kar, a research scientist in the lab of MIT Professor James DiCarlo, drew on previous research. One of the studies he examined involved showing images of faces to autistic adults and neurotypical controls. These images were generated by software that taught them different levels of happiness or fear; the participants were asked to rate whether each face was happy. Compared to the control group, autistic adults needed higher levels of facial happiness to perceive it correctly.

The other study he relied on involved recording neuronal activity in the amygdalas of people undergoing surgery for epilepsy while performing the vision task. This article reported that a patient’s neural activity could be used to predict their judgment of each face.

For the research itself, Kar created an artificial neural network, a computer system that mimics the architecture of our brains and is organized in several computational layers. It trained it to perform the same tasks. The network’s behavior on the emotion recognition task was very similar to that of the neurotypic controls. Then Kar started dissecting it to understand how it did its job and to find clues as to why autistic adults interpret emotions in facial expressions differently than neurotypical individuals.

First, he reports that network responses could be made to most closely resemble those of autistic participants if the output was based on the last layer of the network. This layer most closely mimics the IT cortex and is at the end of the visual processing pipeline in primates, he explains, citing previous research.

Second, Kar looked at the role of the amygdala. Working with the previously recorded data and accounting for it in the output of his network, in which the effect of the IT cortex was already quantified. This showed that the amygdala itself has a very small effect. Together, these two findings indicate that the IT cortex is strongly involved in the differences between neurotypic controls and autistic adults.

He goes on to explain that his network could help select images that would be more efficient for the diagnosis of autism.

“These are promising results,” says Kar. Better methods will certainly be developed “but often in the clinic we don’t have to wait for the absolute best product.”

To validate the findings, he trained separate neural networks to match the choices of neurotypical controls and autistic adults. For each, he quantified how strong the connections between the last layers and the decision nodes were; those in the ‘autistic network’ were weaker than in the network corresponding to neurotypical responses. This, he explains, indicates that the neural connections that interpret sensory data are more “noisy” in autistic adults.

Such a view was further reinforced by Kar by adding varying levels of fluctuation (‘noise’) to the functioning of the last layer of the network that models autistic adults. Within a certain range, this extra noise made the network’s responses very similar to those of autistic adults. Addition to the control network had a much weaker effect in matching responses to those of neurotypical adults.

Although based on how computers work, the findings strongly point us towards answers regarding the differences between data processing in neurotypical and autistic brains.

The article “A computational probe into the behavioral and neural markers of atypical facial emotion processing in autism” has been published published in The Journal of Neuroscience

#understand #autistic #brains #process #faces #differently #artificial #intelligence

Leave a Comment

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