Artificial intelligence created by Brazilian mathematicians prevents pregnancy diseases

A group of Brazilian doctors and mathematicians managed to create an algorithm capable of analyzing MRI images to detect problems in pregnancy and, thus, avoid complications in the fetus and mother.

With a % hit rate when compared to doctors, the technology speeds up the diagnosis of problems in pregnancy and could be implemented in remote parts of the country, which do not have specialists, the researchers say.

Developed by IMPA (Institute of Pure and Applied Mathematics) and by professionals from the Dasa health group , the algorithm is able to identify in the images what amniotic fluid, a substance that helps to protect the fetus and that also collaborates in its development during pregnancy.

Heron Werner, doctor Responsible for fetal medicine at Dasa in Rio de Janeiro and one of the members of the group, he explains that measuring the amount of amniotic fluid is important because it reflects on the well-being of the fetus. The dynamics of the amniotic fluid for more or less can guide us in looking for some pathology in the baby, he says.

The new technology can, for example, help in the diagnosis of obstruction in the esophagus, a problem that prevents the fetus from drinking the liquid, as a result, the amount of the substance in the uterus increases, which can be seen in the MRI, says the doctor. She is also able to perceive dysfunctions in the baby’s urinary system, as this reduces the amount of amniotic fluid that appears in the images.

To reach these conclusions, the researchers used 700 3D magnetic resonance images. They were used to train artificial intelligence, which was further aided by a technology called convolutional neural networks.

“These networks are like a very simple simulation of a functioning human brain inside the computer. It has the artificial neurons and the synapses, which in the brain are really the connections between the neurons, and it simulates these neurons so that the algorithm can learn,” says Roberto Oliveira, a researcher at Impa in charge of the research.

With these neural networks, the algorithm is able to visualize several examples of what it should learn and, from there, it constantly evolves.

Paulo Orenstein, a researcher at Impa and coordinator of the project, says that learning is also due to what the algorithm tries to do. “We guide him about which of these kicks are good and which are bad, and that way, eventually, he starts to learn some ways to detect amniotic fluid in the images,” he explains.

At one point during the research, for example, the algorithm pointed out that an area of ​​one resonance was amniotic fluid. When a doctor analyzed the case, however, he realized that, in reality, what was there was a cyst.

The researcher then warned the machine about the error and she came to understand that areas like that are not amniotic fluid, which has improved her analysis capacity.

For the future, the researchers’ idea to make the algorithm evolve into interpret other organs of the baby as well, such as the brain. But, for this, it will be necessary to provide different data from what is currently used.

Although the technology used in both cases is similar, artificial intelligence needs to learn the specific characteristics of each organ that will be analyzed, explains Orenstein.

When we measure the baby’s head, we also measure the bone . What we want to create is a way to calculate with accurate brain tissue volume. So, [a evolução do algoritmo] would be a more reliable way to measure the development of the baby’s central nervous system, says Werner.

The research is part of the Pi Center (Projects and Innovation) of IMPA, a structure that tries to bring the institute closer to public and private entities.

In addition to the partnership with Dasa, other works developed by the initiative include a project in the Ministry Paraba’s public to detect, through mathematical resources, the manipulation of product prices.

Marcelo Viana, general director of Impa and columnist at Folha, states that there are some reasons for the development of these initiatives, such as returning to society the investments made in the institute, making researchers more professional and raising new financial resources.

The main difficulty is for companies to know what mathematics can do for them and that is why it was important for Impa to adopt this proactive posture [por meio do Centro Pi], says Viana.

Back to top button