Choosing protocols has been a heated controversy for decades in the field of neurofeedback. Ask any practitioner about the best protocols for migraines or ADHD and you get a different answer almost every time. Everybody seems to have a favorite protocol for a given symptom or disorder and many often believe their protocol is best. Some of this has to do with where they received their initial training in neurofeedback. What seems to have an even greater influence on the type of protocols they utilize, is the equipment and software they use. Each vendor derives their software to a large extent from their own neurofeedback paradigm as well as the ways they interact with their end users to further meet the end user needs within that paradigm.
Just as in psychology, where there are 7 basic paradigms (or more), there are several dominant paradigms in neurofeedback and vendors tend to support one paradigm over the other. Some support traditional neurofeedback approaches and others, such as zscore training, are still experimental and highly technical in nature. There are also those vendors and their users who take a medical approach to qEEG and neurofeedback over a behavioral view. These vendors typically have a lot of customers who are medically trained such as neurologists, neuropsychologists or general medical practitioners. The behavioral oriented practitioners on the other hand are counselors, social workers and clinical psychologists. In truth, neurofeedback comes from a psychophysiology perspective which emerged from biofeedback. Consequently, it easily adaptive to both the medical and behavioral perspective.
The medically oriented vendors and their end users tend to be interested in locating Regions of Interest (ROIs) or networks that target specific functions. Neurofeedback is viewed in the same manner as a drug or medical procedure in this case. It conforms to the same policies and concerns that apply to all medical procedures and insurance guidelines. The software they use is very sophisticated and typically requires years of training to fully understand. Mastering this approach requires a high level of technical knowledge in multiple areas including signal processing, statistics, electronics, anatomy and physiology as well as special medical training.
The behavioral or psychologically oriented practitioners tend to be interested in shaping behavior, rather than treatment of disease. They tend to utilize standard neurofeedback protocols along with psychophysiological technologies for support such as HRV and other biofeedback applications. Different standards of risk and benefits inform their practice than apply in medically oriented practices as well as different levels of oversight and conduct. The behavioral oriented software tends to be more basic and accessible and aimed at counselors with masters degrees and minimal technical experience. Their background doesn’t involve training in signal processing, advanced statistics, electronics, medical anatomy and physiology as well as special medical training. The learning curve isn’t as steep and software tends to be more user friendly.
The standard neurofeedback protocols used predominantly by psychologists and counselors have been investigated the most extensively and have the most support from research. This is due to the fact that the medical community has looked upon the technology as less valid than the behavioral oriented community. Other protocols such as zscore training using sLORETA and other methods are even more experimental and while still effective are less supported by research (Coben et al., 2019). They are, however, more attractive due to their apparent sophistication and adaptation of medical language, concepts and graphics. This article, however, is addressed to behavioral practitioners with a different scope of practice than medically oriented practitioners and consequently it addresses the standard more traditional protocols.
One approach to protocol selection is a one size fits all method based on a diagnosis category. These protocols came from early developments in the field but can be extremely effective in some cases. For instance, if a client came with a diagnosis of ADHD, then the practitioner might train beta or SMR up and theta down, preferably at Cz. In most cases this would work with cases of ADHD that involved excess frontal delta, theta and/or deficits of beta. The caveat is that presently we are aware of many subtypes of ADHD that require many different protocols at different locations. Initially, some vendors came up with a fixed list of protocols to use for each disorder indicating a specific placement of the electrode(s). They proposed initiating them at different locations and in specific sequences depending upon a multitude of criteria harvested from clinical observations. In a variation on this some practitioners and vendors also came up with a symptom list that guided protocols and placements to address the symptoms. Some of the protocols were derived from the published research and some of them came from clinical experimentation and observation.
With new information coming in from neuroimaging, locations began to be associated with functional hubs and symptoms were thought to be addressable by training those hubs so that the EEG was more in statistically normal limits. Thus, for speech problems the clinician would look to F7 which correlated with Broca’s area and look for a deviant component band that could be trained to more normal activity. The problem with this approach is that Bocas area was often operating within normal limits with no deviant delta or theta to train down despite the symptoms present. It soon became apparent from further research that networks like speech were highly complex and widely distributed throughout the brain and consequently utilized many locations or nodes. So, the question turned into which node, group of nodes or location would reduce the symptom(s).
Early on, qEEG, and then sLORETA, were thought to be able to resolve the question of location and protocol selection but never did. With multiple locations deviant in multiple neurometric dimensions the problem of selection grew and so did the arguments and conclusions. To make matters worse, many different solutions or strategies actually produced results. Some vendors promoted training all the locations in all neurometric dimensions at the same time. This worked but not any better than other methods (Coben et al., 2019) and placed undue burdens on the client and practitioner. Those practitioners initiated in vendor workshops promoting this approach often were not informed or made aware of the comparable effectiveness of simpler approaches due to vendor bias or commercial considerations.
Training location is important, as Barry Sterman proved when he discovered that training over the motor strip rather than over the seizure focus worked best with temporal lobe-based seizures (Sterman & Egner, 2006). Most of us as clinicians have experience that a change in location can at times make a dramatic difference. As a consequence of this, qEEG, although not required and not ubiquitous, is a common component of neurofeedback today. qEEG does provide advantages and allows the practitioner to get a bigger picture of what is going on. One can readily see significant deviances in magnitude or power as well as other neurometric dimensions such as dominant frequency or asymmetry. In fact, there is so much conflicting information at times that it can overwhelm the practitioner. Thus, new strategies are evolving to deal with this growing input problem including expert programs and AI, but there comes with this the risk of allowing programs to have too much power in the protocol decisions.
At NewMind we have developed a compromise and created a hybrid approach that is effective as well as empirically based but includes and encourages clinical decision making. We statistically determine the most deviant locations in several key neurometric dimensions by rank order and select the top two pairs such as F3-F4 or C3-C4 to train. At each location we identify the frequencies that are deviant in amplitude in order to determine the best filter settings and frequencies to train up or down. We use the quadrant rules published in “Doing Neurofeedback” (Soutar & Longo, 2022) to make this selection. Then we try each protocol and location indicated and observe the results in the trend screen to determine if the component bands are moving toward a normative pattern. We also use a symptom tracker to determine the level of expected subjective symptom changes. This way we have both empirical quantitative and qualitative evidence of the best protocol. Through this method we can identify the most effective protocol with the least side effects. Once the protocol is begun, we monitor progress and change locations when progress plateaus. Then remap after 20 sessions.
The details concerning why we train at homologous sites is in my chapter entitled “Perspective and method for a qEEG based two-channel bi-hemispheric compensatory model of neurofeedback training” in the Handbook of Clinical QEEG and Neurotherapy by Collura and Fredrick (Soutar, 2017). It is based on research concerning the dominant networks and their locations as well as a model of symmetry based on neocortical dynamics. I and many other senior people in the field believe it is unadvisable to train all locations at once because the brain requires room (degrees of freedom) to reorganize and should not be excessively constrained from making its innate required adjustments. Our science at present doesn’t have sufficient information to determine how such a process should unfold in sufficient detail and to determine this with any precision.
Nobody has a crystal ball or magic algorithm to determine in advance what the best protocol for any given individual with any given disorder is going to be. We simply don’t have the science to do that, and I would be wary of anyone who claims they can. The brain is simply too complex, and every individual is too different. It is best to have several probable solutions, preferably statistically determined, and test each one empirically. This way we make the best estimation and waste the least amount of client time and resources.
References
Coben, R., Hammond, D. C., & Arns, M. (2019). 19 Channel Z-Score and LORETA Neurofeedback: Does the Evidence Support the Hype?. Applied psychophysiology and biofeedback, 44(1), 1–8. https://doi.org/10.1007/s10484-018-9420-6
Soutar, R. (2017). Perspective and method for a qEEG based two-channel bi-hemispheric compensatory model of neurofeedback training. In T. Collura & J. Frederick
(Eds.), Handbook of clinical QEEG and neurotherapy (pp. 387-403). New York, NY: Routledge
Soutar, R., Longo, R. (2022). Doing neurofeedback: an introduction. Greenville, SC: Foundation for Neurofeedback & Neuromodulation Research.
Sterman, M. Barry and Egner, Tobias (2006). Foundation and Practice of Neurofeedback for the Treatment of Epilepsy. Applied Psychophysiology and Biofeedback, Vol. 31, No. 1, March 2006.