Doctors measure blood pressure to track heart disease, and scrutinize insulin levels in people with diabetes. But when it comes to depression, clinicians must rely on people’s self-reported symptoms, making it difficult to objectively measure a treatment’s effects.
Now, researchers have used artificial intelligence (AI) to identify a brain signal linked to recovery from depression in people treated with deep-brain stimulation (DBS), a technique that uses electrodes implanted into the brain to deliver electric pulses that alter neural activity. The team reported1 their results on 10 people with severe depression, in Nature on 20 September.
If replicated in a larger sample, these findings could represent a “game-changer in how we would be able to treat depression”, says Paul Holtzheimer, a neuroscientist at the Geisel School of Medicine at Dartmouth in Hanover, New Hampshire, who was not involved in the research.
Table of Contents
Efforts to treat depression with DBS have so far had limited success: two randomized-controlled trials2,3 failed to demonstrate a benefit compared with a placebo. One problem, says Helen Mayberg, a neurologist at Icahn School of Medicine at Mount Sinai in New York City, and a co-author of the Nature paper, is that doctors only have access to self-reported data to assess whether a person’s stimulation voltage needs adjustment.
With self-reported data, clinicians have a difficult time distinguishing between normal, day-to-day mood fluctuations and pathological depression, says Todd Herrington, director of the DBS programme at Massachusetts General Hospital in Boston, who was not involved in the research.
Wiring up the brain to beat depression
To find a more objective measure of depression recovery, Mayberg and her colleagues developed a DBS device that includes sensors to measure brain activity, as well as the standard electrodes for brain stimulation. They implanted this device into the subcallosal cingulate cortex — an area of the brain that has a role in regulating emotional behaviour — in ten people with depression that resisted all forms of treatment.
After 24 weeks of stimulation, nine of the ten participants showed a substantial improvement in their symptoms, and seven met the criteria for disease remission.
Mayberg and her colleagues used an AI model to identify the brain patterns associated with severe depression, drawing on the participants’ brain recordings at the start of the study. They also trained the model to identify the brain patterns that were associated with successful treatment, using the brain recordings from the end of the study. (The researchers obtained usable brain-recording data from only six of the ten participants because of what the study called “prototype device challenges”.)
The model identified a distinct change in neural activity that could distinguish between the two states (depression versus recovery) with an accuracy of more than 90%. All participants had this brain signal associated with recovery. One participant responded well to treatment for four months and then relapsed. After reviewing the participant’s brain-recording data, the researchers found that the ‘recovery’ signal disappeared a month before the relapse. If clinicians had had this data during treatment, they could have changed the stimulation regimen, potentially thwarting the relapse, Mayberg says.
It will be important not only to validate this approach in more people, but also to dig into the biology of what the brain pattern for recovery represents, Herrington says. Mayberg and her colleagues are already testing an updated DBS device on other participants. The new design, unlike the prototype devices used in the research, has approval from US regulators, which will make it easier to launch trials if the results hold up, says Mayberg.
These findings could have implications beyond DBS, Holtzheimer says. People with severe depression might not want an electrode in their brain. But if there’s a corresponding brain signal that could be measured by less invasive methods, the findings might be more broadly useful.