Republish

We encourage you to republish this article online and in print, it’s free under our creative commons attribution license, but please follow some simple guidelines:
  1. You have to credit our authors.
  2. You have to credit SciDev.Net — where possible include our logo with a link back to the original article.
  3. You can simply run the first few lines of the article and then add: “Read the full article on SciDev.Net” containing a link back to the original article.
  4. If you want to also take images published in this story you will need to confirm with the original source if you're licensed to use them.
  5. The easiest way to get the article on your site is to embed the code below.
For more information view our media page and republishing guidelines.

The full article is available here as HTML.

Press Ctrl-C to copy

Scientists are using algorithms, big data and machine learning for disease prevention and control, but support and investment are lacking, writes Martín De Ambrosio.

[BUENOS AIRES] Dominican researcher Rainier Mallol was at NASA’s Singularity University in Silicon Valley, California, when he met his future partner. Malaysian epidemiologist Dhesi Raja told him about how difficult it was for officials to plan dengue control strategies when they did not know when the next outbreak would be triggered.

“I thought it could be a good idea to create a machine learning algorithm to recognise [a potential outbreak] in advance,” says Mallol, who was named one of the top under-35 innovators in Latin America by the MIT Technology Review in 2017. His AIME (artificial intelligence in medical epidemiology) technology is now in use in Malaysia, has been tested in Río de Janeiro and is expected to be trialled next in Sao Paulo.

“Scientific literature says that rain, higher temperature and winds in some regions raise the chances of dengue outbreak. My plan was to trace a map of accumulating data from databases from all over the world,” he explains. “I had 800 variables, including what the houses are constructed of, then reduced it to slightly less than 300.” All that information, he adds, generates an output predicting the likelihood of a new outbreak three months in advance.

“AI is taking its firsts steps, there is a lot to understand and improve. Intelligent systems can’t make a diagnosis better than a physician yet,”

Germán González, National Scientific and Technical Research Council, Argentina

This is one of many projects created by scientists, researchers and entrepreneurs from Latin America using artificial intelligence, including algorithms, machine learning and big data, for different health applications. These range from anticipating disease outbreaks and malnutrition to Alzheimer’s diagnosis and stem cell detection.

These innovators are working in an isolated, uncoordinated way, pressing ahead despite a lack of support from governments and reluctance of big companies to invest in research and development. “There is a lack of coordination and the efforts come from personal interests of the researchers or from some institutes or companies,” says Germán González, bioinformatics researcher from the University of Entre Ríos in Argentina.

The easy and inexpensive way of using algorithms, and the existence of qualified people in the region, means AI has huge potential in the field of medicine.

Predicting dengue, Zika, chikungunya and other mosquito-borne diseases is the goal of another project, from the National University of Córdoba, in Argentina. There, a team led by Juan Scavuzzo uses machine learning to identify in advance the likelihood of the emergence of an Aedes aegypti mosquito colony. It was tested in the northern town of Tartagal, where hot conditions create a good environment for these diseases to spread.

The ultimate goal is a geospatial risk map for the whole country — and not only for dengue and Zika. “Our idea is to continue mixing geospatial data with AI to tackle health issues,” says Scavuzzo. “We are about to publish a paper in Geospatial Health where we show how neural networks can predict Chagas’ disease in rural areas in Argentina."

Predicting problems

Fellow Argentinian Cristian Yones, an engineer from the National University of the Litoral, in Santa Fe, has developed an idea he calls the child growth prediction tree (CGPT). He designed an algorithm which can predict child malnutrition using anthropometric measures such as weight, length and body mass index.

“The system is trained with data of 80,000 children, an amount a paediatrician could never have in mind; we do not aim to displace their position but help them with a great tool,” says Yones.

The award-winning CGPT is about to be tested in the field in Ramón Carrillo Hospital, in Santiago del Estero, before being extended to San Juan and Mendoza hospitals.

These algorithms, if properly designed, could also help predict other medical and biological conditions such as psychiatric disorders and stem cell differentiation.

With just a picture

Santiago Miriuka is not an engineer but a physician. However, he managed to develop AI software that can help him in his daily work in the laboratory.
 
“We wanted to predict stem cell differentiation at a very early stage so we designed an algorithm that can do it with 90 per cent precision in half an hour, and 100 per cent in one hour,” says Miriuka, who is a researcher for Argentina’s National Scientific and Technical Research Council (Conicet) working at Fleni medical research centre in Buenos Aires.
 
His team needs only a picture of the stem cell to obtain the output from the system; in other words, to know what kind of cell that particular stem cell will become – a neuron (nervous system cell) or a hepatocyte (liver cell), for instance. “We can see gene expression in real time with zero costs: only with a picture and running the algorithm,” he adds.
 
The results were published in Cell magazine last March. The aim is to use the application to recognise apoptosis, or cell death. “It has great potential in new drugs trials. AI will change the whole of medicine. Algorithms don’t get it wrong,” adds Miriuka.

A daily dose of AI

AI also has everyday applications for health facilities. Engineer Fredi Vivas explains how his company RockingData processes “ordinary data” to improve health management.

“Our core is not predicting diseases — we are not that glamorous,” he jokes. “We focus on systems efficiency like patient cancellation of medical appointments: we can forecast it as the airline industry does, overbooking with no collateral damage.”

The software also is used to detect fraud in certain health services: appointments billed by doctors but not taken by patients. “The algorithm zooms into the data to recognise the likelihood of a scam and then a human should make an assessment of the situation,” he adds. What does all this mean? Is it going to change medicine in its entirety, as Miriuka predicts? And what is the role of Latin America in the so-called AI revolution?

Germán González, who also works at Conicet, answers: “AI is taking its firsts steps; there is a lot to understand and improve. Intelligent systems can’t make a diagnosis better than a physician yet.

“In Latin America and worldwide, AI is in transition to the real world. Today, most of the developments are in academia and there is a lack of applications. So if Latin America takes some key strategic decisions, we can get, at least, on an equal footing with other regions in the world.”