Can neurofeedback go deeper?

Feedback, a word with more than 100 years of history, is one of the most crucial concepts to understanding how the brain works. Feedback occurs both at the cellular level – neurons are wired to feed information forward and back – as well as at a more behavioral level. We understand feedback colloquially as the reactions we get from the people around us that help to shape our behavior, but the brain uses feedback at multiple levels, down to individual cells and circuits. For example, you use visual feedback to correct the position of your hand if you can not find the light switch at first try, or mostly tactile feedback if the room is dark. At a more cognitive level, negative feedback from people around us is discouraging, while positive feedback encourages us to repeat the same behavior. Inside the brain, one can find signals that reflect something like this higher-level behavioral feedback information from the outside world. One famous example is the dopaminergic neurons in midbrain encoding reward and the lack thereof. The brain, a highly dynamic structure fed by these feedback information, changes in order to optimize actions that maximize reward. This continuous learning process provides adaptation to our changing environment.

To get a sense of what might be happening at the neural level, let's go back to the first example of the motor system's ability to control your arm and hand as you flip a light switch. The motor cortex controls the basics of intentionally moving your hand to the correct location, but if you miss the switch for some reason, the cerebellum takes charge of solving the mysterious discrepancy by comparing outgoing motor commands against sensory feedback, which acts as an error signal. This feedback controller continuously compares the desired output ("hit switch") with the actual result ("there's no switch here!"), and makes adjustments until the desired movement is achieved.

Scientists feed back to the disordered nervous system

Increasingly, clinical applications take advantage of the feedback mechanisms that are extensively used by the brain. The most prominent clinical applications so far are brain-computer interfaces (BCI) using implanted electrodes, such as those used to control neuroprosthetics, and EEG neurofeedback (NFB). BCI is a kind of brain computer interaction where a device provides direct communication between brain and computer. This approach is being used for neuroprosthetics that might substitute sensory, motor or cognitive modalities. There are devices, for example, that assist individuals with quadriplegia. These invasive devices decode the neural activity in motor cortex that are received from implanted electrodes, while the patients think about moving their limbs in a certain way and command a robotic limb that executes the movement. While they get trained for the device, individuals learn to use their neural activity to manipulate computer cursors on screen in order to command the device perform simple motor tasks just by thinking about the task and seeing the visual feedback. Thus, visual feedback provides the cues for the accuracy of the aimed movement, just like the real thing.

EEG NFB uses brainwaves recorded from electrodes placed on scalp and conveys them to a computer, which decodes these signals. The signals are then visualized on the screen so that the individual learns to self-regulate the underlying neural signals picked up by the electrodes. Thus, NFB is distinguished from the BCI approach in several aspects. Unlike the implanted BCI electrodes used for neuroprosthetics, NFB is not invasive, relying instead on EEG electrodes placed on the skull. While implanted electrodes can record from multiple neurons in a very defined region, EEG recordings reflect an aggregate activity, recorded from a rather large, indistinct region, resulting in low spatial resolution. Similarly, temporal resolution of implanted electrodes are better than the EEG electrodes. However, the invasiveness makes neuroprosthetics available only to patients whose conditions are not treatable with any other method, while NFB is basically available to everyone.

BCI and NFB are used for different clinical purposes. While BCI has mainly been developed to assist individuals with severe neurological deficits, ongoing NFB research often aims to ameliorate various psychiatric conditions such as addiction, ADHD, depression, autism and anxiety, besides initial successes with Parkinson’s disease, tremor and dystonia. Neuroprosthetics with implanted electrodes can be considered as a last resort for the patients whose lives are severely affected by the disease, and who cannot benefit from any other therapy. In the case of psychiatric disorders, medication and psychotherapy can generally keep the disease under control, or the patients can live a somewhat normal life without intervention. As a result, individuals with these disorders cannot yet benefit from the more advanced, but also more invasive methods used with BCIs.

In recent years, researchers have begun experimenting on invasive methods for the treatment of pychiatric conditions. One example where implanted electrodes have been used to treat psychiatric disorders is deep brain stimulation (DBS) which has been successfully used to treat major depressive disorder in recent years and found to be effective in many conditions that are otherwise incurable . In addition, there are clinical trials ongoing for DBS treatment of obsessive compulsive disorder. DBS requires implantation of deep brain electrodes as in the BCI approach, but does not require any feedback training as the treatment is comprised of continuous stimulation of an affected brain area.

Neurofeedback for DBS patients

Stimulation of specific brain regions seems to be effective against some psychiatric disorders, but these approaches have not yet harnessed the power of feedback that has been so effective in EEG NFB treatments. Here, I would like to argue that we can, and maybe should, take advantage of these invasive electrodes that are implanted for DBS, and perform neurofeedback. This would be an invaluable opportunity for those individuals to focus on how to feel better (less anxious, less sad, more attentive etc). Currently, DBS is thought to work through altering the abnormal activity in affected regions in a way that decreases symptoms. The patient has no control over the stimulation or its influence on the symptoms. I propose that it might be better to have patients learn to self-regulate this activity in real time rather than just applying standard stimulation. In the development of neuroprosthetics, we have become aware of the importance of feedback to the patient to allow self-regulation. Currently, for instance, a paralyzed patient can move a robotic arm by just thinking about it, but because there is no sensory feedback, the simplest movements are at best slow and clumsy, leaving patients with a lot of frustration. Researchers now are working on providing sensory feedback to make the robotic arm closer to a real one. Similarly, neural feedback can be applied to DBS patients to improve the quality of the improvement.

An important note here is that psychiatric problems are inherently more complex than the neurologic ones such as paralysis in terms of the subjective awareness of the existing problem and capacity to imagine what it would be like to be “normal”. During BCI training, a paralyzed patient needs to imagine herself moving a limb, which is a very easy task. Additionally, research shows that the brain activity is similar when a particular movement is performed vs. imagined. However, let's say for a patient experiencing depression, imagining to be happy or motivated might not be as intuitive. Similarly for an individual with ADHD, being more focused and attentive is a difficult state to attain. Revealing neural correlates of these deficits have thus been crucial in providing the connection between the brain activity and what the patient experiences. NFB shows that patients can self-regulate the brain activity if it is presented to them in the form of a sensory stimulus.

There are two steps that need to be taken to make invasive NFB possible: First is to explore how the neural activity recorded by DBS electrodes correlates with the symptoms of the condition of interest. There are several human imaging and animal electrophysiology studies that begin to address this, but we need to be able to precisely interpret the correlation of symptoms to neural activity in real time. The success of EEG NFB in various psychiatric conditions suggests that this is an achievable goal. Secondly, the electrode placement for invasive NFB might be slightly different than DBS, which would require more research. Additionally, there are technical challenges, such as the need to both stimulate and record from the same electrodes, that will eventually need to be addressed.

Incorporating neurofeedback to DBS not only has the potential to improve the quality of the treatment, but it might give rise to longer term, persistent effects. If the patients learn to control the abnormal activity, or even just get a feedback on how it is changing with stimulation, they might eventually learn to suppress or change the symptoms by themselves. This obviously opens a whole new avenue of research about how plasticity and learning may come about with neurofeedback.

To conclude, it is crucial to benefit from the existing technologies as much as possible. neuroprosthetics, DBS and NFB are rapidly developing, promising techniques and I believe that invasive NFB is worth exploring with the hope of open a new avenue for long incurable psychiatric conditions.