Chapter 11 – Session Management and Control

Chapter 11 – Session Management and Control

In a simple example, if a reward is presented when theta is below some threshold, then the brain will learn to differentiate that state. If it is a beneficial state, and the brain can achieve bxetter balance or focus, then that state change may become learned for future use. However, the key element of the feedback is its contingency, which is to say, it is contingent on some condition. As is true with all learning paradigms, the choice of what makes the reward forthcoming is the crux of the value of the training.

Protocol Selection and Design

In planning a neurofeedback session, the protocol must be defined. This can be as simple as “reduce theta at this location,” or as complex as “train coherence between these two sites up, while reducing beta in another location.” In current technology, it is possible to design protocols such as “normalize connectivity across the rear of the head in the alpha and beta bands while keeping theta and high beta at minimal levels.” All of these protocols define some underlying brain dynamic that will have clinical relevance as it is achieved through operant training.

Whenever using a protocol that appeals to the notions of “larger,” “smaller,” or “within normal ranges,” the concept of a threshold arises. Theresholds are the decision points, and every protocol, no matter how simple or how complicated, appeals to the notion of some type of threshold. The threshold defines the contingency, in that it tells the system when, based upon the EEG parameters, the rewards will be forthcoming, and when they will not.

The use of thresholding is one of the most important and also controversial topics in neural feedback. A threshold consists of a value that is used as a decision point in creating feedback. Thresholds can be set by the manual methods, automatic methods, or a combination of both.

Targets are generally of one of two types. Enhanced, reinforced, or “uptrained” targets are conditions that lead to the possibility of reward. Inhibit, or “downtrained” targets remove the possibility of reinforcement. By using combinations of reinforcements and inhibits, the training can be configured to lead the trainee toward any desired state, or even combination of states. The protocol can be described in terms of the percentage of time that enhance or inhibit conditions are met. The total rate of reinforcement will be the mathematic combination of all of these conditions, which translates into the overall success rate of the training protocol.

The following examples serve to illustrate the importance and value of adjusting thresholds based upon the relevant contingency to be emphasized for operant learning. A critical aspect of learning is that the individual will discern precisely when rewards are forthcoming, and reinforce that state. As in the previously described example of the college professor, an organism does not have to be consciously aware of the learning process. Whatever state or condition is rewarded will tend to be reinforced. Thus the time-specificity of feedback is translated into state-specificity of the learned behavior.

In the first example, which is a typical set of threshold settings, there is one enhance and two inhibits. As is typical, this example uses a low inhibit, such as delta or theta, a midrange enhance such as alpha, low beta, or beta, and a high inhibit such as high beta. If the enhance band is set to reinforce 60% of the time, and the low inhibit is true 20% of the time, and the high inhibit is true 10% of the time, then the total success rate will be 43%. In this case, the inhibits are in effect relatively infrequently, so that the main differential that the trainee will be exposed to is the presence of the midrange enhance. This therefore becomes emphasized in the training.

Threshold example:

• 
  Low inhibit (theta) – 20% TOT
  
  • 
    80% success rate = 100 - 20
    
• 
  Midrange enhance (SMR) – 60% TOT
  
  • 
    60% success rate
    
• 
  High inhibit (Hi Beta)– 10% TOT
  
  • 
    90% success rate = 100 - 90
    
• 
  Expected reward rate:
  
  • 
    0.8 x 0.6 x 0.9 = 0.43 = 43%
    
• 
  Midrange enhance is emphasized
  

In the second example, the target thresholds are changed so that the contingencies are now 80% enhance, 40% on the low inhibit, and 10% on the high inhibit. As shown, the total reward contingency is still 43%. However, the client is now receiving more information relative to the low inhibit, because the midrange enhance is now significantly “easier.” The training experience will weighted more toward discerning the absence of theta, and less toward the production of SMR. This would be a useful strategy to use with a beginning child, for example, whose first task would be to learn to reduce theta, before the production of SMR becomes a the focus of training.

In this final example, the thresholds are further adjusted so that the enhance band is above threshold 100% of the time. The net result of this maneuver is to essentially ignore this band. It is thus removed from the training experience. Rather, the theta inhibit is set for 60% time above threshold, and the high inhibit is set to never be above threshold. The total reinforcement rate is still 43%. However, the training experience is now entirely that of theta inhibition.

These examples serve to illustrate how neurofeedback can be configured so as to “titrate” the balance of training conditions. In a manner analogous to the use of combinations of medications in approaching medical treatment, neurofeedback can be configures specifically, to target particular brain states and operant behaviors. In this manner, even though the underlying mechanism is quite simple, that of modulation of the concentration/relaxation cycle in the brain, the configuration of the neurofeedback protocol introduces significant target specificity that can be used to reinforce particular brain states, and even locations, via. selective feedback.

In the same way that these adjustments lead to specificity in time, they can also lead to specificity in space. For example, if the system is configured to only reinforce alpha at a particular location, such as the back of the head, then the client will learn to produce alpha in that location. The ability to select location is not unlike that of a strobelight, which is a light that turns on at particular times, to reveal a moving object only at specific times. A strobelight can be used to “stop” a moving wheel, by flashing only when the wheel is in a particular position. Similarly, if the neurofeedback protocol is configured to reinforce only a particular location and condition, then the brain will learn to reinforce that condition. When used in conjunction with advanced localization techniques such as LORETA or sLORETA, neurofeedback can produce very specific training effects, which are reflected in localization in space, as well as in time.

It is important to recognize the role of the inhibit bands in ensuring that the reinforced overall state is appropriate. In early EEG biofeedback devices, only the reinforced band was monitored. Typically, the individual was rewarded whenever signals in a filter set to the alpha frequency band exceeded a threshold level. A significant concern arose in that if the trainee did anything to raise this level, including blinking eyes, moving, or creating muscle tension, then a reward would be forthcoming. This made it possible to reinforce these actions, as well as actual alpha activity. Until inhibit bands, also known as “guard” bands were introduced, this problem could not be overcome, and results of “alpha” training were inconsistent.

Another strategy is to use relative power rather than absolute values, in determining feedback. In this case, there is an automatic inhibition of signals outside the target band. Another way to look at relative power is as “percent energy” for a band. For example, if the percentage of alpha is rewarded whenever it exceeds 40%, then the client will experience rewards when their alpha becomes 40% or more of the power. One advantage of this simple approach is that, if the client produces out-of-band energy such as eyeblinks or muscle tension, this will tend to reduce the percentage of alpha accordingly. Therefore, the client will not receive rewards when these confounding signals are present, and the effect will be to inhibit any out-of-band activity, without the need for explicit inhibits on these frequency ranges.
The question inevitably arises of when and how to adjust thresholds. This question is as old as operant training itself, and has fundamental importance. Unfortunately (or fortunately), there is no simple answer. This question is as complicated as the entire topic of learning, and any method that introduces contingency is a feasible approach. As it turns out, a wide range of philosophies and practices are in use, and will likely continue into the foreseeable future. This reflects the art of neurofeedback as well as the importance of individual differences between clinicians, as well as between clients.

Because neurofeedback is based upon the trainee’s experience, it has elements of art as well as of science. This is particularly relevant when determining the strategy used to set thresholds, as these decisions effectively shape this experience. While certain fundamental rules of operant learning are at play, there is still flexibility in particular decisions, and these will depend to some extent on the preferences of the clinician as well as the client, and the particular goals. The possible strategies and their considerations are summarized in the text box. To summarize briefly, it is possible to justify the adjustment of thresholds from one extreme, which is to fix them without change throughout a training program, to the other extreme, which is to adjust them continually.

If thresholds are set once at the beginning of training, and never adjusted, there is the benefit that the trainee will see the effects of their improvement over time, and be rewarded for progress. One disadvantage is that the training may become too easy over time, because the targets are not adjusted in response to improvement in performance. One solution to this is to move to the next level, which is to adjust them once for each session. In this case, the client can be told of the threshold changes, such as “today we will use a target of 6 and see how you do.” It would be important in this case to tell the trainee where the thresholds are set each day, as part of the reward process. Some practitioners, however, want to emphasize the use of an “optimal” rate of reward, and adjust thresholds every few minutes. In this case, the feedback can also include a periodic display of the progress in terms of the component values themselves, so that the trainee see their own progress. Finally, some practitioners will arrange for thresholds to be adjusted continually, so that the client is informed whenever the signals go “above where they have been recently,” arguing that this will reward any improvement, even over short periods of time.


When to adjust thresholds?

• 
  Never
  
  • 
    Don’t frustrate trainee
    
  • 
    Allow to see improvement in scores
    
• 
  Once for each session
  
  • 
    Tell trainee new threshold
    
  • 
    Goal of consistent number of points per session
    
• 
  Every 2-5 minutes
  
  • 
    Optimal rate of reward
    
  • 
    Show trainee improvement in EEG scores
    
• 
  Continually
  
  • 
    Brain is a dynamical system
    
  • 
    Provide information regarding emergent variability
    

Clearly, both extremes of this decision process have disadvantages, which are reduced in the middle ground. Therefore, adjusting thresholds for a particular session, or possibly re-adjusting them periodically, tends to be more common. It should also be noted that threshold adjustment, particularly when it is done automatically (“autothresholding”) is controversial, and some practitioners are strongly against this practice. It can be argued that if thresholds are changed too often, then the trainee is training the biofeedback system, not the other way around.

Sessions can be continuous, lasting for 10, 20, or more minutes, or they can be broken into “trials” or “runs.” These are often separated by brief pauses or breaks. These are, again, individual decisions that will depend on the training and experience of the clinician, as well as the particulars of the task. For example, alpha training for deep relaxation is often conducted in a single, continuous session, facilitating the achievement of the relaxed, internalized state. SMR or beta training, on the other hand, tend to have a “task” aspect, and can be punctuated by periodic pauses every 2, 3, or 5 minutes, during which the trainee rests and reviews session progress.

Another consideration in neurofeedback is how many sessions to use, and how long sessions should last. These are again a very individual decisions, based upon a variety of factors. Such issues as seriousness of the problem, availability of the trainee, and travel issues may come into play. As a general rule, neurofeedback will be done no less than once a week, so that learned gains can be reinforced and retained. At the other extreme, training more than once or twice a day, day after day, represents an extreme. This can be used if a client has to travel a long distance for neurofeedback, and is coming in for an intensive treatment program. In cases of difficulty reaching the clinician, one option is to consider home training, but only after several in-office sessions, and appropriate training of the parents or other family members who will be assisting the client.

Sessions generally last less than an hour, with the exception of when alpha/theta training is used to achieve deep states, and is used in conjunction with psychotherapy, guided imagery, or related interventions. It is typical for the duration of actual neurofeedback training to last for a minimum of about 10 minutes, and a maximum of about 30 minutes, in most clinical applications. It is often possible to determine the optimal session length by monitoring the progress of the relevant EEG variables, as well as the client’s subjective report. In particular, if the EEG values start to deviate, indicating fatigue or loss of connection to the learning goals, or if the client reports that they are getting tired or bored, training should be discontinued for that session.

The number of sessions used will vary. Some clients may report results in less than 5 or 10 sessions, and wish to move on to other therapies, or consider their treatment done. Others may require 40 or more sessions, depending on the severity and permanence of their condition. When neurofeedback is used in pervasive conditions such as Autism Spectrum Disorder (ASD), continued neurofeedback may be indicated, with the number of sessions reaching or exceeding 100.