Scientists have proposed a method for predicting the level of intelligence of children by MRI of the brain using three-dimensional networks of deep learning.
Magnetic resonance imaging isA common way to capture images of internal organs. The researchers decided to study the possibility of using the brain images thus obtained to determine the biological intellectual potential of adolescents, regardless of knowledge, age and gender.
For this, a team from Skoltech created a three-dimensional neural architecture network which to increase the accuracy of the forecastapplied several mathematical models to the same data at once. For the analysis of MRI using machine vision, a new ensemble method was used, which allows you to classify images entirely and without loss of information.
The study found a correlationbetween the level of "fluid" intelligence of the child and the anatomy of his brain. Although forecasting accuracy is still below the desired level, the developed models open up new aspects of children's development.
Machine learning technologies also allowidentify signs of illness in children, which even doctors may not pay attention to. For example, researchers recently developed a phone app that analyzes photos to look for early signs of eye disease.