When parents learn about Michael Milham’s research, they often ask him, “Can you give my child a brain scan to figure out what’s wrong with them?”
Milham treats his young patients like any other child psychiatrist would: He observes and interviews them, assigns them diagnoses, and prescribes courses of treatment. But unlike many psychiatrists, Milham is also a scientist—he is vice president of research at the Child Mind Institute—and an expert on functional magnetic resonance imaging, or fMRI, a tool that allows researchers to measure levels of activity across the brain.
He understands why parents want him to scan their children’s brains. For families in search of an explanation for their child’s distress, the inexactitude of psychiatry—its overlapping diagnoses, its uncertain prognoses—can be frustrating. “There’s no other part of medicine that behaves this way,” Milham says. “Every other part of medicine has laboratory tests.” More than a century into its modern existence, psychiatry still lacks an objective method for determining precisely what is causing someone’s distress and how best to treat it.
So scores of scientists, including Milham, are working to change that. They are combing through the genomes, brains, and immune systems of tens of thousands of study volunteers in search of psychiatric “biomarkers”—measurable, objective indications of someone’s psychiatric state. For some, the dream is something like a brain scan for diagnosing autism, but many scientists are doubtful that such goals will ever be achieved. “[Brain imaging] is not ever going to be, in my opinion, useful for telling somebody whether or not they have a problem,” says Desmond Oathes, assistant professor of psychiatry at the University of Pennsylvania.
That doesn’t mean, though, that fMRI can’t do anything for psychiatry. Even if a brain scan can’t tell someone what condition they have, Oathes says, it could one day help them recover. “It can be super, incredibly useful for saying which treatment is going to be most effective.”
fMRI is a younger cousin of the more-familiar structural MRI—the sort of scan you might get if you come into the hospital with an injured knee. Whereas structural MRI maps out the locations of different body tissues, functional MRI detects levels of oxygen in the blood. Working cells need oxygen, so a high blood oxygen level suggests that a brain region is active. Scientists can analyze how these activity levels change over time in a single brain region, or use them to infer the strength of the connection between two regions.
When fMRI emerged in the 1990s, it revolutionized the field of human neuroscience. For the first time, scientists could watch the entire human brain at work without injecting subjects with a radioactive agent, which previous technologies had required. But although fMRI has become a bread-and-butter technology of neuroscience, it is rarely used in medical settings.
Neurosurgeons might examine their patients using fMRI to map out different regions of their brains before operating, but psychiatrists aren’t sticking anyone into a scanner.
Which is not to say that fMRI has nothing to do with psychiatry at all. Since the invention of fMRI, scientists have been scanning mentally ill people’s brains to try to figure out what’s different about them. Reams of scientific papers have associated psychiatric diagnoses with brain regions that are bigger or smaller, more active or less active, more strongly or less strongly connected. And some of these findings are quite robust: Numerous studies have shown, for example, that the amygdala and the insula, two brain regions associated with emotion, become overactive in depressed people when they experience something negative.
These are important results. They can fuel theories about the neurological causes of depression, and they may indicate which regions psychiatrists should target with brain stimulation therapies. But they won’t help Milham’s patients receive a diagnosis by brain scan. Just like knowing that men are taller on average doesn’t allow you to guess someone’s sex solely based on their height, amygdala activity levels won’t help you figure out who is depressed.
Psychiatric biomarkers just aren’t precise enough for diagnosis, and experts say that’s down to the way that those diagnoses work. “They’re meant for reliable communication between clinicians,” Milham says. “They’re not meant for valid research.” There just might not be any observable differences between the brains of people with anxiety and those of people with depression, and that means fMRI is the wrong tool for the job.
The tried-and-true, low-tech clinical interview is simply better suited to the task at hand, says Dani Bassett, professor of bioengineering at the University of Pennsylvania. “We don’t need fMRI—we don’t need any biomarkers actually—to do diagnosis.” But unlike in other areas of medicine, where a diagnosis (e.g., type 1 diabetes) indicates a solution (insulin), in psychiatry a diagnosis is just the beginning of a treatment journey. Different drugs and therapies work better for different parents, and clinicians don’t know ahead of time which options will prove most helpful. And because it can take months for psychiatric treatments to show their effects, patients may continue to suffer for a long time before they and their clinicians hit on the optimal treatment.
So scientists are trying to use brain scans as a tool for predicting which treatments might work best for specific patients. Various studies have used fMRI to try to distinguish patients who will respond to antidepressants from patients who won’t. Most promising, according to Milham, are those studies that leverage fMRI to optimize one of the buzziest new depression treatments: transcranial magnetic stimulation (TMS). “That’s the low-hanging fruit,” he says.
To the uninitiated, TMS can seem like magic. A technician waves a plastic-covered, figure eight-shaped wand over the right spot on a patient’s head, and the patient—who hasn’t responded to multiple other treatments—recovers from their depression. But it’s real, and it works; the FDA approved TMS as a depression treatment almost 15 years ago. The wand generates a magnetic field that changes the level of electrical activity at a specific spot in the brain, and for many people, that stimulation can alleviate their debilitating mental illness. A typical course of treatment involves around 20 or 30 half-hour treatments delivered over the course of a month, and some patients remain in remission a full year after TMS treatment.
But not everyone. From the earliest studies of TMS in depression, patients have fallen into two categories: responders and nonresponders. Some see marked improvement in their mental illness, and others experience no change at all. Given the way that the treatment is typically administered, Oathes says, this divergence in response isn’t surprising. To try to find the target site for TMS treatment—the left dorsolateral prefrontal cortex (dlPFC), a brain region associated with cognitive abilities like decision-making and memory—technicians will first identify the patient’s motor cortex by passing the magnetic coil over the surface of the patient’s skull until they move involuntarily. Then, they shift the coil five centimeters forward and assume that this location is the dlPFC.
On some people, though, the dlPFC doesn’t lie precisely five centimeters in front of the motor cortex. “It was a guesstimate, a very rough guesstimate,” Oathes says. So in a recent study, scientists at Stanford didn’t just make scalp measurements in a blind search for the dlPFC; instead, they took fMRI scans of each subject’s brain in order to precisely identify the specific subregion of dlPFC that they wanted to target. More than 90 percent of subjects achieved remission immediately after treatment, and most remained in remission a month later. The treatment was atypical in other ways—patients were given TMS 10 times a day, far more frequently than is standard, so the impressive recovery rate might not be fully attributable to fMRI-based targeting. But Oathes sees the results as a promising indication of how fMRI might eventually make its way into the clinic.
There are other potential applications of fMRI in TMS treatment, especially when it comes to regions that lie deep within the brain. Areas like dlPFC can be directly targeted by TMS, because they sit at the brain’s surface. But the magnetic fields generated by the TMS can’t reach as far as the amygdala and insula. The only way to affect these regions with TMS is to do so indirectly, by stimulating a region at the surface of the brain to which they are strongly connected. fMRI could help clinicians monitor whether TMS stimulation is successfully reaching those deep regions, and the strength of connections between different regions might even let them predict which patients will respond to TMS.
Given the state of the research, experts think that brain scans could start to be used more widely in TMS treatment sometime in the next decade. But moving brain scans from research settings to the clinic isn’t a trivial task. Structural MRI scanners can be programmed to take functional MRI measurements, but the raw data collected by such machines needs extensive analysis to be rendered usable. That analysis takes two things: enormous computational resources and substantial training for the people who do it.
For now, clinicians might outsource the analysis to researchers. But fMRI tends to attract a lot of hype—hence the parents asking Milham to scan their children’s brains—and Oates worries that it could compromise fMRI’s potential to improve psychiatric treatment. If the use of fMRI in psychiatry expands too rapidly, clinicians who haven’t received specific training in fMRI data analysis might try to use it to optimize treatments themselves, without the support of researchers. And this unjudicious use of the technology could undermine the progress researchers have labored for years to achieve.
“I don’t want people doing that,” Oathes says. “It scares me to death.”