Francisco Pereira was one of the speakers at the Responsible AI Forum 2025, held in Lisbon on November 25, where he discussed artificial intelligence (AI) in the healthcare sector.
In cases such as depression, “technology can assist researchers in understanding why one patient differs from another, based on data collected about the patient such as medical history, clinical tests, and brain images,” said the head of the Machine Learning Core department at the National Institute of Mental Health (NIMH), USA, the largest funder of mental health research worldwide.
“However, we currently lack the means to use this information to predict which medication will aid each person; this remains a significant area of investigation. With the availability of various data types on patients, there is hope that we can use this to help personalize treatment for each individual,” explained Francisco Pereira.
“Our greatest assistance to our colleagues involves efforts like quantifying patient-reported elements,” he added.
For instance, “we work with colleagues” who “aim to help parents communicate better with children undergoing depression treatment.”
It involves situations like, “parents may complain: You’re too passive, you do nothing. And the child responds: You don’t understand what’s going on in my head,” and colleagues, he noted, “have a therapeutic sequence that works to help parents and children communicate better in this context, aiming to understand what is happening as treatment progresses.”
The objective “is to assist our peers in better understanding what aspects relate to the ongoing treatment,” he added.
Regarding AI tools in the healthcare sector, there are numerous applications, yet “many have never undergone formal evaluation to determine whether they count as alternative treatments and function better,” pointed out Francisco Pereira.
On using AI, “my perspective is that of a scientist aiming to aid health researchers,” helping scientists “comprehend why a treatment is effective.”
“Our laboratory operates on that basis,” he emphasized, highlighting that a large part of their work involves analyzing data that “scientists manage to collect,” whether it’s questionnaires, MRI scans, or other forms, and “trying to assist people in creating better models of what they’re studying.”
“Our group’s main focus is to utilize machine learning and artificial intelligence or other probabilistic, mathematical models, allowing scientists to pose questions they otherwise could not,” he mentioned.
When asked what goal he aims to achieve, he states definitively: “Helping colleagues studying irritability in children predict which factors are most likely” to cause loss of control and irritability, including shouting, breaking things, among others, referring to Disruptive Mood Dysregulation Disorder (DMDD).
“The aim is not to predict that the child has the disorder, as the problem is evident; it’s more about placing colleagues in a position to assist parents,” because “we have models that, based on questions regularly answered via app, can predict with some accuracy if there is a higher likelihood of the child losing control soon,” he explained.
This can be used to remind parents of the techniques learned from psychologists to manage such situations.
Regarding research aims for the coming year, Francisco Pereira expressed a desire to “see this clinical trial functioning,” as they are still developing the pilot intervention.
“Our group, Machine Learning Core, conducts data analysis and application projects for over 40 laboratories at the NIMH. We also undertake our own developments in machine learning and statistics to address new questions” within the institute and “provide consulting and training for researchers throughout the NIMH.”
