I’m lying in the plastic cocoon of an MRI machine, an instrument that measures activity in different parts of the brain. As I try to hold still, the loudly clanking machine runs a structural scan to locate the anterior cingulate cortex and the insula, regions involved in processing pain. A computer then translates the MRI signal into three small animated fires, representing the activity levels of the cingulate and the right and left insula, projected onto a screen above my face.

I concentrate to make those fires roar and ebb, using only my thoughts. As I do, the MRI is measuring changes in the blood flow to selected parts of my brain. The patterns of blood flow tell the computer how neural activity is changing. By trying to control the size of the fires, I am attempting to control brain activity in the cingulate and insula, and in turn to quell the chronic back pain that has irked me in recent years.

Monitoring my progress is Christopher deCharms, a neuroscientist and founder of Omneuron, a startup company in Menlo Park, CA. DeCharms has spent the last five years developing imaging techniques that can be used to teach patients to control their brain activity. Changes in neural activity usually take place unconsciously, as different parts of the brain are engaged to perform tasks or process stimuli. Neurons in the language circuit start firing, for example, when you have a conversation with a friend. When you watch a scary movie, neurons in the amygdala, an area involved in emotion, fire more frequently. But consciously controlling these changes – damping activity in specific brain regions – could theoretically be useful for treating not only pain but such diseases as depression or even stroke. Exerting that kind of control is difficult, but it may offer an alternative to drugs that is both more precise and less likely to cause side effects.

Until a few years ago, selective control of brain activity was just a provocative idea. But a new version of functional magnetic resonance imaging (fMRI) has, for the first time, made brain activity visible in real time. The technology was just what deCharms needed. He and his collaborator Sean Mackey, associate director of the Pain Management Division at Stanford University, have already shown that their technique works, at least in the short term. In December, they published the results of their first study in the journal Proceedings of the National Academy of Sciences, showing that both healthy subjects and chronic-pain patients could learn to control brain activity – and pain – using real-time fMRI.

“There are potentially dozens of diseases of the brain and nervous system caused by an inappropriate level of brain activation in different areas,” says deCharms. He cautions that fMRI feedback is not yet ready for clinical use – he and Mackey are still confirming their results in long-term clinical trials. But even as he refines the use of the technique for treating pain, deCharms is now testing it in patients with anxiety disorders. And other scientists are running or planning pilot studies of fMRI feedback to treat depression, stroke, attention deficit hyperactivity disorder (ADHD), and post-traumatic stress disorder.

Mind Control

DeCharms was still a graduate student at the University of California, San Francisco, in the 1990s when he started studying how the neural connections in the brain grow and change with experience, a phenomenon called neuro-plasticity. Neuroscientists knew that repeatedly exercising parts of the brain can elicit permanent changes in the complex neural circuitry responsible for, say, hearing or vision. DeCharms theorized that by consciously increasing or decreasing the neural activity in specific brain areas involved in disease, patients could control some of their symptoms and perhaps permanently change their brains for the better. DeCharms believes that patients with depression, for example, might be able to use fMRI feedback to learn to control the neurons that release the signaling molecule serotonin, and perhaps the cells serotonin acts on, as well. This would achieve the same goal as drugs like Prozac – increasing the amount of serotonin available in the brain – but might not produce side effects.

“If you practice a new form of dance, the first thing that happens is you learn to do the activity better. You engage the musculature, and it becomes stronger,” says deCharms. “Eventually, your physical body has been changed. It’s a long-lasting effect, even when you’re not consciously trying.” One key to strengthening the right dance muscles, of course, is feedback on your performance: dance studios always have mirrors on the walls. DeCharms hoped the same process would work in the brain, if he could find a way to measure brain activity rapidly and accurately enough for patients to learn to control it and to mimic desired patterns.

The idea of using feedback in the brain is not new. For 30 years, scientists have used electroencephalograms (EEG) – a technology that measures electrical activity coming from the brain – to train people to elicit or maintain a particular type of electrical pattern. Results from preliminary studies suggest that such training is somewhat effective for treating ADHD and substance abuse, though large, placebo-controlled studies have not yet been completed. But because EEG technology picks up electrical activity spanning multiple brain areas, its usefulness for specific feedback is limited. DeCharms wanted to target the anatomically tiny brain structures involved in disease, and in sensations like pain.

In contrast to EEG, fMRI measures the blood flow in precise areas of the brain, yielding much finer spatial resolution. It shows which areas are working hardest during a specific task, and it can also point out which parts of the brain are functioning abnormally in specific diseases. But for deCharms, it was the development of real-time fMRI that was the breakthrough. FMRI generates an enormous amount of data, which used to take days or weeks to analyze and interpret. But newer algorithms and greater computing power have collapsed that processing time down to milliseconds. That means scientists – and subjects – can watch brain activity as it happens.

For deCharms and his collaborators, this type of fMRI held a powerful appeal. They theorized that people with neurological or psychological disorders could perform mental exercises to try to modulate activity in specific neural systems that had gone awry and get immediate feedback on which strategies were most effective. Then they could use those strategies to feel better.

Tigers and Pain

I’ve suffered from chronic back pain for five years, the symptoms persisting despite an array of treatments: stomach-wrenching amounts of ibuprofen, prescription painkillers that made me woozy, lengthy ergonomics consultations, and months of physical therapy and acupuncture. My problem is not uncommon. An estimated 50 million Americans suffer from chronic pain, and for a large percentage of those patients, existing therapies are inadequate.

Pain is a complex phenomenon. It depends both on neural signals that are generated during tissue damage, as when you grab a hot pan, and on a higher-level system that interprets those signals to form the pain experience – an interpretation that may be altered by your emotions and level of attention. For example, soldiers wounded on the battlefield often don’t feel the extent of their injuries until they are out of danger. So while pain is an adaptation that evolved to help us avoid bodily injury, our brains have also evolved a sophisticated system for turning it off. “You need to be able to run from a tiger, even if you’re hurt,” says deCharms.

DeCharms chose pain as his first test of real-time fMRI technology, partly because the need is so great and partly because the neurological circuit that underlies pain is well understood. Opioid drugs, such as morphine, target these neurons chemically. Implantable stimulators, which can be an effective treatment for pain, target the circuit with small jolts of electricity. DeCharms, on the other hand, wanted to try to target the system consciously, through cognitive processes.

In last December’s paper in the National Academy journal, deCharms, Mackey, and their collaborators described a study in which participants learned a series of mental exercises derived from strategies used in pain clinics. For example, they might have been asked to imagine the sensation of their brains’ releasing painkilling compounds into the aching area, or to imagine that their painful tissue was as healthy as a pain-free part of their body. Subjects then climbed into the MRI scanner, where they wore special virtual-reality goggles that displayed the activity in a part of the brain involved in feeling pain – the anterior cingulate cortex. They were instructed to try to increase or decrease the activity by performing the exercises. The MRI data gave them direct feedback on how well their mental strategies were working, allowing them to adjust their technique. Some people picked up the knack quickly, while others needed several sessions to learn appropriate control methods.

Eight patients with chronic pain that wasn’t adequately controlled by more conventional means reported a 44 to 64 percent decrease in pain after the training, three times the pain reduction reported by a control group. Those who exercised the greatest control over brain activity showed the greatest benefit.

The researchers also designed an elaborate set of controls to show that the results didn’t simply reflect the placebo effect or an artifact of the experimental process. For example, subjects who did not get fMRI feedback but were instructed to focus attention to and away from their pain did not show as much pain relief. Patients who got fMRI feedback from another part of the brain also did not benefit; nor did patients who got feedback from the cingulate of another person. “If expectation or being in the scanner were contributing … then that group should have seen a similar result,” says deCharms. The researchers also conducted tests in which chronic-pain patients were given more-traditional biofeedback data, such as heart rate or blood pressure. Patients who received fMRI feedback had a significantly greater reduction in pain.

However, some scientists say it’s still not clear what kind of role attention, or even emotion, is playing. “In our experience, people are so engaged in the task, they don’t even know how long they’re in the [MRI],” says Seung-Schik Yoo, a Harvard University neuroscientist who is also studying real-time fMRI. “If someone is so captivated, they could forget to pay attention to the pain.” And success in controlling the activity levels shown on the screens could further distract a patient from the pain. “When it works, time flies,” says Yoo. “When it doesn’t, you get frustrated.” He adds that the best way to determine whether test subjects are permanently affecting their brains will be a long-term clinical trial, like the one deCharms and Mackey have under way. Still, says Yoo, “Their work has paved the way in pain control using this new technique.”

All in Your Head

When I told my father about my trip to Omneuron, he asked a question that deCharms is asked often. If you can mentally control pain, why do you need MRI feedback? Shouldn’t the pain, or lack of it, be feedback enough?

The short answer is no. “No other technique that involves feedback has been able to do this sort of thing that well,” says Peter A. Bandettini, director of the fMRI core facility at the National Institutes of Health in Bethesda, MD. According to Bandettini, figuring out why the fMRI feedback is effective is one of the big remaining tasks. He says the answer lies partly in the way fMRI pinpoints precise areas of the brain. But that still leaves a huge question: how do patients actually manipulate the activity in those areas? How do they will control over activity levels? “People figure out how to change the activation, but they don’t know exactly what they do,” he says. “I think if we learn more about that, the technique will become more widely applicable.”

Mackey hopes to eventually unravel the neural systems responsible for the painkilling effects. It’s possible that activating the cingulate leads to the release of chemicals such as endorphins, natural painkillers produced by the brain. In fact, the process may be similar to the one that causes the placebo effect. Placebo treatments can have a profound effect on pain and on certain diseases, notably depression – even inducing changes in the brain. Recent studies show that sham painkillers can trigger the release of endorphins and activate the anterior cingulate, the same brain area under scrutiny in the feedback study. According to deCharms, fMRI feedback may provide a way to consciously control this process.

Even if they are uncertain about the mechanisms behind fMRI feedback, biomedical researchers are excited about exploring its possibilities. “The results from deCharms’s experiment are compelling enough that people will probably be jumping in,” says Bandettini. Adds Tor Wager, a psychologist at Columbia University, “The field of neurofeedback is wide open. … We need more research that explores what people can do themselves.” The possibilities are likely to grow as neuroscientists zero in on the brain areas responsible for different functions and the specific abnormalities linked to different disorders.

Many experts caution, though, that it’s still too early to determine the broad therapeutic potential. “We’re going to have to do the studies and see if feedback is helpful,” says John Gabrieli, an MIT neuroscientist who collaborated with deCharms and is now planning to test fMRI feedback for ADHD. “We need to figure out which disorders are amenable, how long the effects last, and what contexts are needed to support them.” And as in any test of a novel technology, the findings must be repeated in other labs.

It’s possible that some parts of the brain are more susceptible to conscious control than others, and such differences could limit the number of areas that are responsive to fMRI feedback. The anterior cingulate cortex, for example, may be easier to control because it is involved in attention, which we actively modulate throughout the day, as we work or daydream, read or watch television. Diseases such as depression or social phobias, which can often be treated effectively with behavioral therapy, might also be good candidates for fMRI feedback, says Gabrieli.

Yoo, meanwhile, hopes to show that fMRI feedback could speed rehabilitation from stroke or other brain injuries. Patients often lose a particular function, such as speech or part of their vision, when such an injury kills a cluster of neurons. Sometimes the brain can heal itself, either spontaneously or through practice, by reorganizing nearby neurons to take over. This process generally takes place unconsciously, but Yoo says fMRI feedback could teach patients how to consciously activate the regenerating areas.

Among the most compelling therapeutic possibilities is a combination of fMRI feedback with cognitive behavioral therapy, a popular form of talk therapy in which patients learn to change negative thought patterns. During a standard session, a patient might tell the therapist about an event that provokes anxiety and then use specific mental exercises to calm down. In the version deCharms and colleagues are testing, a patient lies in the scanner and communicates with a therapist in the next room through a speaker. Both therapist and patient can watch the patient’s brain activity throughout the session. Using that information, patients might try to consciously alter the activity patterns that flare up when they become anxious.

A Painful Lesson

Before I hit the scanner at deCharms’s lab, we practice a few of the mental exercises that he routinely teaches his subjects. I imagine my brain releasing endorphins, their painkilling signals traveling down the length of my spinal cord to reach my lower back. To try to increase my pain, I imagine that my lower back is burning. (Trying to worsen the pain sounds counterproductive, but deCharms theorizes that learning to modulate pain in both directions will give patients more power over brain activity.) I’m shocked by how sharply I can make the pain flare up.

Now that I’m inside the scanner, the screen instructs me to try to increase or decrease the size of the fires representing my brain activity. I set to work, trying to focus simultaneously on my pain and on the screen overhead. The fires wax and wane a bit, sometimes smoldering, sometimes burning at a steady pace. My pain that day is mild, and it’s difficult to tell if the fires are flickering randomly or at my will. Try as I might to extinguish the flame or coax it to a roaring blaze, the fire mostly burns low.

After about 15 minutes, the technician’s voice crackles over a speaker in the scanner – my first session is over, and to my surprise, I did achieve some control. She projects onto the screen a rough graph comparing activity in the cingulate during the intervals when I tried to increase the fires with the activity when I tried to decrease them. There is a clear difference between the lines.

When the technician asks if I want to try another session, I agree, determined to do even better this time around. During this session, I switch mental strategies, which deCharms recommends as a way to find the technique that works best. Instead of imagining endorphins being released in my brain, I focus on the healthy tissue of my hand and try to imagine that my back feels just as pain-free. The fires on the screen flicker and flare, and I’m convinced I have a better handle on my neural activity. When I receive my official results several weeks later, I discover I was right. I performed best during my last session, successfully controlling the activity in my right and left insula.

DeCharms is now trying to determine the best ways to teach fMRI feedback; if long-term studies confirm his team’s initial findings, and the U.S. Food and Drug Administration approves the treatment, he eventually hopes to open treatment clinics. Like a complex dance, the technique is hard to pick up, and some people are naturally better at it than others. “We need to figure out who is good at this and how to make it easier,” says deCharms. His team is developing new ways to display brain activity to make feedback more effective. The fire graphic used in my session, for example, is a relatively new addition. The researchers are also doing extensive psychological screening to see if people who easily learn to control their brain activity have identifying characteristics. One of the biggest factors will probably be motivation. Feedback somewhat resembles exercise, albeit an odd mental form of it – so it requires willingness and effort.

My own test run is just a single afternoon, and I can’t tell if my pain is any better. But I did seem to control select parts of my brain. And for better or worse, after two hours in the scanner, I am definitely conscious of my lower back.

Emily Singer is the biotechnology and life sciences editor of Technology Review.