Antidepressants are prescribed using a trial-and-error approach. Experts hope that artificial intelligence can change this.
Artificial intelligence (AI) can help psychiatrists predict whether antidepressants will be effective, which could save patients time and improve their psychiatric care, according to a new study.
Although antidepressants may be prescribed to treat moderate to severe depression, healthcare providers should wait six to eight weeks to see if there is improvement and adjust treatment if there is not.
Using artificial intelligence, researchers at Amsterdam University Medical Center (Amsterdam UMC) and Radboud University Medical Center in the Netherlands have found a way to reduce this delay.
They focused on sertraline, also known by the brand name Zoloft, which is one of the most prescribed first-line treatments for depression.
“For half of them it works, for the other half it doesn’t. It means that many weeks are wasted, and with the accumulation and with withdrawals, the cycles as we call them, can take up to six months,” he told Euronews Maarten Poirot, PhD student at UMC Amsterdam and first author of the study.
The team brought together different data, including MRI predictors, such as hippocampal volume and blood flow, and clinical information, and developed an algorithm for this.
“We found that integrating all this information can lead to a clinically valuable model for predicting outcome at eight weeks,” Poirot said.
“The good thing is that the study is very informative and contributes to the critical need to develop accurate machine learning (ML) methods that can guide treatment decisions. This is a critical need especially among patients with mental illnesses,” Dr. Soroush Saghafian, an associate professor at Harvard University who was not involved in the study, told Euronews.
Depressive disorder is estimated to affect around 6% of the EU population and is a leading cause of disability worldwide. World Health Organization (WHO).
“Artificial intelligence is fundamental”
The study, which involved 229 patients aged between 18 and 65, was published in the American Journal of Psychiatry.
“AI is critical as our work is on the border between radiology and psychiatry and until recently, all work in the radiology department was done by people literally looking at images,” Poirot said.
“With the amount of data we acquire and its complexity, this simply wouldn’t work anymore. Also, the patterns can be very subtle and very complex,” she added.
According to Saghafian, the combination of MRI data with clinical parameters is another strength of the study.
“In recent years, many AI and ML algorithms are trained on multimodal data, thanks to the availability of such data, and this has enabled higher levels of prediction accuracy to be achieved,” he said.
According to the researchers, the developed model was able to predict whether the treatment would work in just one week.
“The algorithm suggested that blood flow in the anterior cingulate cortex, the area of the brain involved in emotion regulation, would be predictive of the drug’s effectiveness,” Eric Ruhé, a psychiatrist at Radboud University Medical Center, said in a statement.
“And at the second measurement, one week after onset, symptom severity was further found to be predictive,” he added.
One limitation of the study is that the data has not been externally validated, but the researchers hope to run a clinical trial with unused data to train the algorithm.
“If we could also show the same kind of performance on this external validation, that would further strengthen the confidence we could have in such an algorithm,” Poirot said.
For Saghafian, “the accuracy of reported predictions is relatively low, calling into question the suitability for actual clinical implementation.”
Another limitation of the study is that it focuses on a single antidepressant.
“In reality, patients often undergo a combination of treatments and therefore, to further ensure suitability for implementation in clinical practice, it may be necessary to develop AI and machine learning methods capable of taking into account a combination of treatments and allow us to predict counterfactual outcomes,” explains Saghafian. she said.