Mexicans create AI platform that predicts cases of disease

As their final graduate course project, Javier Rebull and Juan Pérez, students at the Tec's Mexico City campus, along with Luis Sánchez from Laguna campus, developed Epi-forecast MX, a platform that predicts incidences of Parkinson's, Alzheimer's, and depression in Mexico.

The Master's in Applied Artificial Intelligence (MNA) students worked in collaboration with the Mexican Social Security Institute (IMSS) and used data retrieved from the National Epidemiological Surveillance System (SINAVE) to train different models and predict how many cases may arise in a year.

"It's a useful tool for medical institutions such as the IMSS to identify when cases of mental illness increase or decrease and, based on this, to plan and provide better services by allocating resources and adapting consultations," explained Luis.

The AI platform allows users to visualize predictions by state and gender through interactive dashboards: a tool that centralizes real-time data, allowing it to be filtered, explored, and analyzed by users.

It also includes a specialized chatbot. "It's used to make natural language inquiries based on the information we feed into the models," Luis remarked.

The name Epi-Forecast MX refers to "Mexican epidemic forecast." It is a system based on time series models.

These models are tools used to analyze chronologically ordered data and predict future behaviors based on past trends.

The team tested more than 1,000 models and finally chose around 300, which were precisely distributed.

"Each state, gender, and ailment generates a different model, which produces many combinations. "So, what we did was look for a way to automate the training," Javier said.

"No artificial intelligence model is exact: there will always be some error, but what we're looking for is to have the smallest possible error," Juan explained.

With the aim of making the page accessible to everyone, the team explained that they focused on a simple and easy-to-access interface.

"A lot of technical detail isn't necessary for the end-user. This platform aims to allow even non-technical personnel to access the information and use it to support their decision-making," Luis explained.

"It's information that currently exists, is verified, and can be used and utilized today; for me, that's the great value of this project," said Juan.

Today, the team is preparing for a meeting with Health Secretary David Kershenobich in which the goal is to explore the possibility of expanding access to new data sources. This will feed the model with other sources of information and evaluate its potential.

"This will allow Mexico to allocate its public health resources more efficiently, avoiding waste and enabling better year-on-year planning," Javier concluded.

Throughout this project, the team received advice from Grettel Barceló, a tenured professor at the MNA, as well as support from Ruth Pérez at the IMSS and Lina Díaz from the Ramón de la Fuente Muñiz National Institute of Psychiatry (INPRFM).

Barceló explained that this platform could allow for public policy decisions to be made based on projections and evidence.

"It can also be used to see where there is gender or regional inequality in care," Barceló concluded.

DON'T GO WITHOUT READING:

TecSalud receives two international awards in cardiology

Eva Trujillo and her medical struggle to treat eating disorders

What is measles? Symptoms, prevention, and about the health alert