The use of artificial intelligence in mental health could significantly improve the diagnosis and treatment of mental health disorders, and in the near future, an assessment for depression could include a quick brain scan to identify the best therapy and customize it based on the patient’s needs. According to a new study by Stanford Medicine researchers, the combination of brain imaging and machine learning can identify subtypes of depression and anxiety.
About 30% of people with depression have what’s known as treatment-resistant depression, meaning that multiple types of medications or therapies haven’t improved their symptoms. And in up to two-thirds of people with depression, treatment fails to fully reverse their symptoms to healthy levels.
This is partly because there is no reliable way to know which antidepressant or type of therapy might help a particular patient. Medications are prescribed on a trial-and-error basis, which can take months or years to find one that works, if ever. And spending so much time trying treatment after treatment, without experiencing relief, can make depression symptoms worse.
Better methods for matching patients with appropriate treatments are desperately needed, said senior study author Leanne Williams, a professor of psychiatry and behavioral sciences and director of the Center for Precision Mental Health and Well-Being at Stanford Medicine. Williams, who lost her partner to depression in 2015, has focused her work on precision psychiatry.
“The goal of our work is to figure out how to get it right from the start,” Williams said. The study, published in the journal Nature Medicine, categorizes depression into six biological subtypes, or “biotypes,” and identifies treatments that are more or less effective for three of these subtypes.
Response to treatment according to the type of depression or anxiety
To better understand the underlying biology of depression and anxiety, Williams and his colleagues assessed 801 participants previously diagnosed with depression or anxiety using imaging technology known as fMRI to measure brain activity. They scanned the volunteers’ brains at rest and while they performed tasks designed to test their cognitive and emotional function. The scientists focused on brain regions and the connections between them that were already known to play a role in depression.
Using a machine learning approach known as cluster analysis to group the patients’ brain images, they identified six distinct patterns of activity in the brain regions studied. The scientists also randomly assigned 250 study participants to receive one of three commonly used antidepressants or behavioral therapy.
Patients with one subtype, characterized by overactivity in cognitive regions of the brain, responded better to the antidepressant venlafaxine compared with those with other biotypes. Those with another subtype, whose resting brains had higher levels of activity among three regions associated with depression and problem solving, showed symptom improvement with behavioral therapy. And those with a third subtype, who showed lower levels of resting activity in the brain circuit that controls attention, were less likely to see symptom improvement with behavioral therapy.
“As far as we know, this is the first time we have been able to show that depression can be explained by different disruptions in brain function.”
The biotypes and their response to behavioral therapy make sense based on what is known about these brain regions, said Dr. Jun Ma, a professor of medicine at the University of Illinois Chicago and one of the study’s authors. The type of therapy used in their trial teaches patients skills to better deal with everyday problems, so high levels of activity in these brain regions may allow patients with that biotype to adopt new skills more easily.
As for those with lower activity in the region associated with attention and engagement, Ma said it’s possible that a pharmaceutical treatment that first addresses that lower activity could help those patients benefit more from behavioral therapy. “To our knowledge, this is the first time we’ve been able to show that depression can be explained by different disruptions in brain function,” Williams said. “In essence, it’s a demonstration of a personalized medicine approach to mental health based on objective measures of brain function.”
In another recent study, Williams and his team showed that using fMRI brain imaging improves their ability to identify individuals who are likely to respond to antidepressant treatment. In that study, the scientists focused on a subtype they call the cognitive biotype of depression, which affects more than a quarter of people with depression and is less likely to respond to standard antidepressants. By identifying those with the cognitive biotype using fMRI, the researchers accurately predicted the likelihood of remission in 63% of patients, compared with 36% accuracy without using brain imaging. That improved accuracy means providers may be more likely to get treatment right from the start. The scientists are now studying new treatments for this biotype in hopes of finding more options for those who don’t respond to standard antidepressants.
Different biotypes also correlated with differences in symptoms and task performance among trial participants. Those with overactive cognitive brain regions, for example, had higher levels of anhedonia (inability to feel pleasure) than those with other biotypes; they also performed worse on executive function tasks. Those with the subtype that responded best to behavioral therapy also made errors on executive function tasks, but performed well on cognitive tasks.
One of the six biotypes discovered in the study showed no noticeable differences in brain activity in the regions analyzed compared to the activity of people without depression. Williams believes they probably haven’t explored the full range of brain biology underlying this disorder: their study focused on regions known to be involved in depression and anxiety, but there could be other types of dysfunction in this biotype that their imaging didn’t capture.
Williams and her team are expanding the imaging study to include more participants and want to test more types of treatments across the six biotypes, including medications that have not traditionally been used for depression. Her colleague Dr. Laura Hack, an assistant professor of psychiatry and behavioral sciences, has begun using the imaging technique in her clinical practice at Stanford Medicine through an experimental protocol. The team also wants to establish easy-to-follow standards for the method so that other practicing psychiatrists can begin implementing it.
“To really advance the field of precision psychiatry, we need to identify the most effective treatments for patients and get them on that treatment as soon as possible,” Ma said. “Having information about patients’ brain function, particularly the validated signatures we assessed in this study, would help determine the most precise treatments and prescriptions for individuals.”
