Point vision: neural network will detect developmental abnormalities in children under three years of age

Russian developers have created a hardware and software complex with elements of artificial intelligence, which makes it possible to identify developmental problems in children under three years of age. According to psychologists, the technology can help reduce subjectivity in the diagnosis of developmental delays, but it will be possible to trust the results of the proposed test only if its relationship with the child's speech and communication abilities is proven.

How to identify developmental abnormalities in children

Specialists of the Russian National Research Medical University named after N.I. Pirogov have created a hardware and software complex that allows them to objectively identify developmental abnormalities in children under three years of age. Normally, the ability to distinguish oneself from the outside world is formed already in a year and two months. If this does not happen, it may indicate mental problems. The methodology has already passed the state registration procedure and can be applied in practice. The developers emphasize that their proposed approach should not replace a specialist's diagnosis. And the test results should only be the basis for a more in-depth neuro-psychological examination.

It is very difficult to identify the norm and abnormalities of mental development in children under three years of age. This is usually possible only with very severe abnormalities, such as genetic ones. International scales based on parental observations are used to identify less pronounced disorders, so the results depend very much on interpretation. And there are practically no objective methods, the developers said.

— An essential indicator that allows us to talk about the norm of development is the isolation of oneself from the outside world. This is what is called self-identification. Normally, this occurs in children aged from one year two to three months to one year eight months to two years. Using computer testing technology, we have proposed a technique that allows us to do this not "by eye", but to record it in a measurable form," said Vera Nikishina, head of the Department of Clinical Psychology at the University of Pirogov.

For the system to work, the computer screen displays an image of a child obtained with a video camera, with painted marks on his face. During the examination, the mark is placed first on his forehead, then on his nose and finally on his chin. There are three color and size options for each placemark position, which change one another sequentially.

The subject is positioned in front of the screen and the video camera at a distance of 40-60 cm without additional instructions. If the child is distracted, you should pay attention to the screen. If there is no reaction within 20 seconds, the size of the label changes, and then its color changes. When a reaction occurs, the size and color of the label changes to the next one.

— The child looks at his image and sees this point. The camera captures his reaction. If, for example, he brushes this point off his forehead or chin with his finger, then there is self-identification. That is, a marker indicating the norm of mental development has been formed in the subject," Vera Nikishina said.

Such an examination takes only a few minutes, and the system is very easy to use. Experts have already identified the regulatory testing indicators that are typical for children with healthy development. They will be used as reference points for different patient groups. Work is currently underway to assess the response to markers in children with autism spectrum disorders, epilepsy, and intellectual disability.

Early detection of abnormalities in children

The hardware and software complex can have various modifications. For example, you can use not just dots as a label, but an image of a butterfly. Now the program provides the results of the study for conclusions to a medical psychologist. However, the creators plan to train artificial intelligence to do it on its own.

The proposed system evaluates a very narrow indicator, which makes it difficult to draw conclusions about the child's level of development. The lack of the necessary reaction can be explained by the individual rate of development. If even the level of speech development cannot be unambiguously assessed, then it is all the more difficult to do this by the presence or absence of self-identification, says child psychologist Elena Gorlova.

— The main indicator should not be this test, but the development of interaction with objects, interaction with adults and speech development. How quickly the child grasps the logic of action with objects. It is important whether the child has onomatopoeia or individual words at 18 months. Therefore, it is necessary that the proposed methodology be compared with other development indicators. If a correlation with a low level of speech development, communication, and subject activity is identified and confirmed, then a simple testing procedure could serve as an indicator for specialists," the expert said.

The computer system is based on the classic method of checking a child's self-identification: if he tries to remove the mark seen on his face in the reflection, it means he is aware of himself. Normally, this skill is usually formed between 15 and 24 months, although individual variations are possible, explained child psychologist, fairy-tale therapist Daria Dugentsova.

— The novelty of the system is that it objectively records the child's behavior through video recording and computer vision algorithms, reducing the risk of subjective errors by parents or psychologists. The absence of a reaction after 20 months or two years is only a signal for an in–depth diagnosis, but not a diagnosis. It is important to consider the technology in conjunction with other methods and observations of specialists," she said.

As data accumulates, the system can become a useful tool for studying the development of children with different disabilities (autism spectrum, speech delay, etc.). The more statistics collected, the more reliable and reliable the technique for early detection of abnormalities, the specialist believes.