Artificial intelligence is used to diagnose autism – 07/13/2023 – Equilibrium

The diagnosis of ASD (autistic spectrum disorder) is still a great challenge due to the degree of complexity involved, requiring highly specialized professionals. The condition is a neurodevelopmental one whose associated symptoms vary considerably. The incidence, according to the latest report by the Center for Disease Control and Prevention in the United States, is one case in every 36 people and there is no biochemical marker that allows it to be accurately determined.

Proposing a quantitative methodology for diagnosis was the suggestion given by researchers in an article published in Scientific Reports. The work was based on brain imaging data from 500 people, 242 of which belong to the spectrum.

The idea was to obtain the diagnosis with the aid of machine learning techniques based on data from brain networks.

“We began developing a method by collecting these images using magnetic resonance imaging or electroencephalography”, says Francisco Rodrigues, a professor at the ICMC-USP (Institute of Mathematics and Computing Sciences of the University of São Paulo), in São Carlos, and one of the authors supported by FAPESP. “By comparing maps of people with or without ASD, we found that it would be possible to make the diagnosis using this methodology.”

The researchers fed an algorithm, which is a machine learning method, with these “maps”. From the learned examples, the system was able to determine which brain alterations were associated with the spectrum. The accuracy of the tests was greater than 95%.

Many recent works propose methods for diagnosing ASD based on machine learning. However, they focus on just one statistical metric, without taking into account the organization of the brain network, which is the differential of this study, highlight the authors. The brain map, also called the cortical network, shows how the regions of the brain are connected. These networks began to be studied about 20 years ago and have offered a new insight into neuroscience.

“Just as a highway with interruptions alters the traffic in a region, the brain with alterations leads to changes in behavior”, says Rodrigues.

From the images obtained in the MRI scans, it is possible to verify changes in certain regions of the cortex, as well as the fact that the brain networks of patients show more segregation, less distribution of information and less connectivity. These regions are involved in cognitive, emotional, learning and memory processes.

“Until a few years ago, little was known about the alterations that lead to the appearance of symptoms. Today it is known, for example, that altered brain circuits in patients with ASD may be related to some behaviors. However, most anatomical studies shows that the changes are barely visible, which indicates the difficulty of diagnosing the milder cases. Therefore, our work is an important step towards new methodologies that allow a better understanding of this neurodivergence”, says the professor.

The methodology is still under development and will take years to be implemented. However, it is an important step towards determining the differences in the brain and may, in the future, be used to assist specialists, especially in cases where there are doubts regarding the patient’s diagnosis.

Miscellaneous Applications

Rodrigues points out that the work is still a small step towards understanding how ASD is related to changes in the brain and that many studies are still needed for this automatic diagnosis methodology to be put into practice. Brain mapping is an important step towards not just identifying autism, but also other conditions. Previous works show that these maps can be used for the detection of schizophrenia also with great precision.

“A decade ago we started developing new methods for identifying mental illnesses and we found that schizophrenia is a disorder whose diagnosis can be greatly improved with the use of this technology, which uses brain networks and artificial intelligence. Alzheimer’s case and we verified that an accurate automatic diagnosis is possible”, mentions Rodrigues, referring to a study published in the Journal of Neural Engineering in 2022.

There are still many challenges involved, as the databases are small and their collection is not simple. But, as a general methodology, it can help in understanding different conditions – and one of the objectives of the group of researchers is to see the relationship between them.

“How similar, in terms of brain changes, are schizophrenia and Alzheimer’s? If we can link the disorders, perhaps we can develop new drugs and similar treatments for different conditions, or even tailor treatments from one condition to another. , but what is to come is quite promising”, highlights the scientist.

The researchers hope that by better understanding how changes in brain circuitry influence behavior, it will become possible to treat patients more humanely and efficiently, better directing public policy.