Page 1, 2, 3
We are at the threshold of revolutionary change in our approach to mental dysfunctions as they come to be seen primarily in the frame of neural network relations. To date the recovery of brain function has been approached mainly in the frame of neurochemical models, but even this principally involves neurochemistry in the service of neural communication. The medications in common use for mental dysfunctions by and large target the neuromodulator systems that regulate synaptic excitability. But there is another aspect to the problem. Information transport via the action potential mechanism is subject to tight timing constraints, and this represents a major potential failure mechanism for the brain under duress. This is a particular issue following physical or emotional trauma, or when the brain has been diverted from its proper developmental pathway early in life. In these cases there may be little or no evidence of structural injury to the brain. The deficits must lie almost entirely in the functional realm, and therefore should be accessible to a functional remedy. And yet we know that pharmacotherapy is largely ineffective in application to these conditions. The deficits are more easily understood in the bioelectrical domain of timing and frequency. And when remedies emerge based on an appeal to neural network functioning in the frequency domain, the case for a new departure in understanding mental dysfunctions is consolidated. This is where we now find ourselves.
Progress in brain imaging over the last two decades has brought these issues to the fore. At first we had to make do with the static imagery of PET and SPECT, but this already drew our attentions back to the intact, functioning brain, and elevated our gaze from brain slices in petri dishes and the firing patterns of individual neurons. Such a divide-and-conquer strategy gave us no hope of ever fully understanding the living brain. A complementary systems-level perspective was missing. With the emergence of functional magnetic resonance imaging in the early 1990s, we had our first opportunity to witness brain dynamics, albeit with very poor time resolution as the process being monitored was glucose uptake. This led in time to the identification of our core connectivity networks, which constitute the foundation of a hierarchy of functional organization in cortex.1 Follow-on studies revealed that identifications could be made between various mental disorders and dysfunctions in our intrinsic connectivity networks (ICNs).2 On that basis, Menon proposed that much of psychopathology could be grounded in the failure of our core control networks to coordinate properly. Evidence was cited for several conditions, including schizophrenia, depression, anxiety, dementia, and autism. But the mechanism undoubtedly has more general applicability.3 Yet other studies demonstrated that conditioning experiments targeting localized features of the fMRI could elicit functional improvement.4 As an example, deCharms demonstrated benefit for complex regional pain syndrome (CRPS1), yet another relatively intractable condition that does not yield to pharmacotherapy.5 Jointly these findings implicate functional connectivity of the core networks as a basic failure mechanism in mental disorders. The full exploitation of these new findings call for an integrative, systems-level perspective on brain function, one that takes both perspectives – the neurochemical and the bioelectrical domains – into account. And on the clinical side that implies the need for an integrative approach to remediation as well.
Very few scientific novelties burst upon the scene without obvious antecedents, and the same holds true here. Much of what is now being proved out in fMRI-based neurofeedback has been previously demonstrated using feedback with the EEG as a training variable. For example, just 2 years after the publication by deCharms a similar report was published using EEG-based training with CRPS1, and the results were comparable, if not better.6 The method used was just 1 year old at the time. It was infralow frequency training, the principal topic of this article. However, the antecedents of this method go back more than 30 years.
The EEG was our first noninvasive technique for studying brain dynamics, but its complexity limited its clinical utilization, so the world moved on. It has taken new developments in mathematical analysis and in data acquisition to restore the EEG to its proper place in research. This is also having beneficial fallout for clinical practice, with innovation increasingly occurring at the pace of software development. The history of EEG biofeedback, now commonly termed neurofeedback, is elaborated in a newly published book titled Restoring the Brain: Neurofeedback as an Integrative Approach to Health. The book covers recent developments, with particular focus on the rapidly emerging technique of infralow frequency training. The case is made for neurofeedback to play a complementary role in integrative medicine.7
The EEG gives us insight into the brain's regulatory activities and how these are organized both spatially and temporally. The argument briefly goes as follows: Cortex is organized for massive parallel processing in the service of pattern recognition. The organization of such patterns must be supported by neuronal assemblies of macroscopic spatial scale, and these must be organized for simultaneity in processing. The principle that "simultaneity defines belonging" is known as time binding.8 Seen in its simplest mathematical terms, the neuron is organized for coincidence detection – on the timescale of milliseconds – by the nature of the action potential mechanism. Coincidence at the level of the neuron in turn supports simultaneity at the level of the neuronal assembly. Since action potentials are perishable entities that are not self-replicating, the persistence of states must be explicitly organized through repetition. Refresh is supported via a recursive cerebral architecture, thus generating a periodicity that maintains continuity. Simultaneity of action at the level of the neuronal assembly, in combination with periodicity, translates to local synchrony at the level of the EEG. Communication between cortical sites is reflected in synchronous (or at least coherent) activity at those sites. Thus, the regulatory role of neuronal assemblies is reflected in frequency-based organization of sufficient magnitude to become discernible to us in the EEG, giving us exquisite insight into timing relationships at the network level. Hence the EEG distills for us the aspect of neuronal activity that is involved in neuronal regulation (as opposed to information transfer). We don't get to see the text, but rather only the context.
We know that the brain is a highly competent, if not overeager, correlation detector, and in feedback this quality is put to good use. If the brain is allowed to witness selected features of its own EEG, it has no difficulty recognizing its own agency with respect to that signal. Once that loop is closed, the brain inevitably also takes responsibility for the signal and tries to control it. This process is analogous to the brain's taking responsibility for keeping the bicycle properly balanced and the car properly pointed down the freeway. The rider or the driver doesn't have to be engaged on these matters. They can be left to the brain to handle in background. The same principle holds in the case EEG feedback and fMRI feedback. The process does not necessarily involve conscious mediation. All that is required is that the process not be interfered with. Distraction is the bane of the bicyclist, of the driver, and of the neurofeedback trainee.
When the brain is involved in feedback under conditions in which it is allowed to exercise its discretion, there is a bias in the direction of better regulation. Left to its own devices, but with the benefit of physiologically relevant information, the brain tends to move toward calmer, better-controlled states. It is residence in calm, de-stressed states that presents the therapeutic opportunity for the functional reorganization of the core networks. The brain is entirely in charge of the recovery process that follows. We have simply arranged to provide a propitious context. The client is usually aware of nothing more than the feeling of migration to a state of calmness with which he may not have been previously acquainted. (Even squirrely autistic children may adopt a quiescent, meditative pose while undergoing the training.) The client may also remark about a state of alertness that does not seem congruent with such a state of placid calmness.
What has been described above is the typical experience of someone undergoing neurofeedback in the infralow frequency region of the EEG. There are many other ways of doing neurofeedback, however, and some history needs to be reviewed to provide context for the discussion to follow. EEG biofeedback got its start in the 1960s with the training of the cortical resting rhythms, principally the famous alpha band at nominally 10 Hz,9 and the sensorimotor rhythm at nominally 13 Hz.10 The alpha band signal at the occiput could be regarded as the resting rhythm of the visual system, and the sensorimotor rhythm (SMR) could be understood as the resting rhythm of the motor system. Since the activity in these bands has an episodic, bursting character, it was trained in an operant conditioning paradigm of challenge (with respect to a threshold) and reward. In both instances, the objective was to move the trainee toward lower levels of arousal; that is, toward calmer states and improved levels of regulation. In the case of the alpha training, this was found to be ameliorative of anxiety states, and the SMR-training was found to be stabilizing against motor seizures and even temporal-lobe seizures, among others.11,12 These initial findings were vigorously challenged at the time, but they have been amply validated since.13,14 A meta-analysis of the work with epilepsy has also been published.15
The research on SMR-training for seizure management became the springboard for application to hyperkinesis, which later became ADHD.16,17 All but one of the ADHD studies that also tracked IQ found significant improvements in IQ score with the training. A 20-point IQ improvement was observed in two 8-year-old mildly mentally retarded twins. The results were published after 5 years of follow-up, over which time the gains held but no further improvements could be elicited.18 The work with SMR-beta training led to much broader application to psychopathology by the early 1990s.19 It emerged that the challenge to a single dominant rhythm of the EEG in a training paradigm sufficed to evoke broad functional reorganization of the cerebrum.
Subsequent to the early research, alpha training evolved in three directions. One aimed toward optimum functioning through the broadening of the attentional repertoire via the promotion of whole-brain alpha synchrony.20 Another used intensive alpha synchrony training in support of psychotherapy and personal transformation.21 The third and most common application was oriented toward the promotion of deep, internally focused states that facilitated the resolution of traumas and a reprieve from addictive propensities. This came to be known as Alpha-Theta training once theta-band reinforcement was added to the protocol.22-24
The next major thrust in the technology was toward the individualization of what came to be known as SMR-beta training (because reinforcement of the slightly higher-frequency beta 1 band (15–18 Hz) was also commonly included in the protocol). The personalization of the training took two forms. The first of these followed immediately upon the availability of affordable full-brain mapping capability on PCs in the early 1990s. This allowed the transient excursions into dysregulation to be detected wherever they occurred on cortex, and the training strategy to be adapted to focus on such excursions. The training brain was cued with respect to such excursions in what was referred to as inhibit-based training. This involved nothing more than the transient withholding of rewards. Brain activity was not actually being inhibited. The other thrust, for which our own group was responsible, consisted of the individualization of the reward frequency, which began after 1995. It was observed that sensitive and unstable brains were differentially responsive to different EEG frequencies, and that for optimal outcomes the target frequency sometimes had to be tuned to within 0.5 Hz or even less.
Such frequency specificity could be readily confirmed because of its repeatability. In the highly responsive individuals at issue here, we were dealing with demonstrable real-time control of their physiological state. For instance, one person might feel hungry at one frequency and yet experience satiety at a nearby frequency. The therapist could toggle the frequency back and forth and reliably reproduce these feeling states. Another person might feel tearful at one frequency (without having any basis for being teary) and yet feel placid or even upbeat at an adjacent frequency. The transition period could be as little as a minute or two. Yet another person might feel the onset of a migraine aura at one frequency, and observe its subsidence at an adjacent frequency. Again this held true reproducibly. Unsurprisingly, bipolar individuals are particularly responsive, and in some cases can be actively moved between depressive and manic states within a period of minutes with a slight shift in reward frequency. In between these two frequencies lies the optimal target frequency at which the brain can train itself toward stability simply by virtue of lingering there for a number of training sessions. This gives the neural networks the opportunity to consolidate the new configuration.
The implications of this responsiveness are huge. We were in fact effecting real-time alteration of physiological state in the general case. Trainees mainly differed in their awareness of such state shifts. Those with the most sensitive, unstable, or reactive brains were the best reporters on their own state as well as being the most sensitive to the choice of target frequency. For both of these reasons, such sensitive individuals were the canaries that guided further development of the method. In the event, we were led to provide training at ever lower frequencies to accommodate them. By 2006, after several years of gradual migration to lower frequencies, this trend caused us to enter the infralow frequency region. Training was performed at frequencies of 0.1Hz and below, the very region in which the intrinsic connectivity networks were first identified using fMRI. In the following, this will be referred to as ILF training.
At such low frequencies, threshold-based training was out of the question. The cycling time was too slow. We had to resort to signal-following, in which the brain is simply exposed to the exceedingly slow undulations of the signal. The brain's interest was captured by the continuity of information flow, and the whole training exercise became more effective than it had ever been before. For the first time, the brain was being trained in accord with its own preferences rather than being regimented like one of B. F. Skinner's pigeons. The brain was in charge of its own journey, based on its own assignment of meaning to the signal that it was observing. (The reader's indulgence is requested for the use of such anthropomorphic language with reference to the brain as an autonomous agent.) The results were so dramatic – and so parametrically specific – that we promptly issued a Protocol Guide to our practitioner network.25
Page 1, 2, 3