As I have noted before (https://metapsychology.net/poc/view_doc.php?type=book&id=4805&cn=458), neurofeedback, a form of biofeedback, has a checkered history with often a dearth of evidence leading to accusations of charlatanism if one practices it. The brain is a noisy electronic signaling hub, and so we amplify and filter the signals into bins, most commonly Alpha, Beta, Delta, Theta and Gamma, and make attributions about what this means, having to match what are called EEG or rather QEEG signatures to common symptoms and disorders. There then follows, on that basis, a series of interventions based on components that are the result of filtering, resulting on quantitative qEEG analysis with such metrics as coherence, asymmetry, power ratios and phase for instance. All of these require explanation for the uninitiated. Experts like Martijn Arns in Nijmegen in Holland have provided some pivotal data, especially in treating ADHD, but only get one mention here, although he gets more attention elsewhere. One of the Othmers also gets a single mention, which shows how the field has added data and publications in the face of the ‘explosion’ in tech since the first edition.  Dozens of figures pepper the pages, as is needed in a training manual.

The book, as a professional series, has a didactic approach, starting with the basics: What is an EEG feedback practice, and how long, how much etc and other questions to be answered?  Described as experiential learning, I guess operant conditioning is a term that others more commonly use, EEG uses computer-guided learning to achieve its goal and involves classical generalization, extinction and discrimination as Demos explains.  As an intervention, we usually refer to it as temporal and spatial, in that it happens rapidly in the exact moment the brain wave being measures gives rise to the signal on the EEG visualizer. 

Demos describes the product of brainwaves drawing on Edmonds (2015). Brain waves start in the thalamus, communicating in cyclic fashion with the cortex, these rhythms regulated by thalamic pacemakers, unless inhibited in the brain stem by the reticular formation, resulting in EEG desynchronization or lower amplitudes. If not, then cycles of synchronization and desynchronization remain under thalamic pacemaker control. Neuronal pathways form column that resonate with each other slowly (delta or theta for instance), or more quickly as in beta activity. Training alpha for instance during neurofeedback can help to restore the thalamic-brain stem communication. He notes carefully that brain waves are the product of excitatory and inhibitory postsynaptic potentials and not the action potentials themselves.

Once filtered and amplified, the abovementioned shapes emerged, the first simply being Alpha in respect of its shape. This morphology defines what these are called, and the reference to observed human behavior related to that shape on the raw EEG.  Awake and mentally active, beta, awake but at rest, alpha, asleep, theta, deeply asleep, delta, and so on including subdivisions such as HiBeta or Alpha 1 and 2, SMR etc. Filtering is by frequency, and amplitude, and training the EEG is not based on the raw data, which needs to be visually inspected for artifact. This latter concern does bedevil automated qEEG reporting as such programs may not deal well with forehead muscle contractions or the like. Which filters you use determines a lot of the interpretation.

My friend Arns in Nijmegen has told me he won’t interpret other’s qEEGs for instance, and often remarks on the interpretations that I send him for critique, that he personally could and would choose not to extract what some practitioners extract, and so the field is wide open to what practitioners think they can extrapolate from such inspections of filtered and smoothed EEG. Of course, I often send MRI scan reports for second opinion, and this can be a shocker too, with reports that up to 30% of radiologists across the world may not interpret MRI’s accurately, and one to the other, the interpretations can be disparate. Recent criticism has labelled vast percentages of neuroscience publications as borderline nonsense, and sadly psychological research is bedeviled by a lack of replication. So: although this particular field of neurofeedback is heavily criticized, and while Demos notes he has found support for the evidence behind neurofeedback overwhelming, it really is up to the skill of the individual, and years of the right experience, as he notes early in the book.

He covers basic terminology related to the spatial arrangement of the EEG sensors, and anatomical nomenclature. Asymmetry of signal based on localization and type of wave would then lead to some aspects of diagnosis, using statistical operations, such as Z-scores related to standard deviations from normative measures, and then the filtering and modelling that creates the qEEG 2-D brain maps. High Z scores imply the statistical distance from the mean is becoming significant (red on the maps), or less so, (blue on the maps).

As with MRIs that I noted above, to validate the interpretation of the scorer, there needs to be collateral, namely from clinical evaluation using baselining tests, so the EEG is not the gold standard. In typical EEG training circles, the doctrine would be that we treat patients, not EEG’s, and so we rely on clinical signs and symptoms with hopefully confirmatory imaging, such as EEG. Power (amplitude) training and Z-Score training of identified sites in the EEG would then follow, if all roads lead to Rome by a careful series of triangulation of measures and symptoms.

Descriptions of mono and bipolar placements of single-channel amplifiers are then given, and reward and inhibit protocols are then described. This means rewarding the trainee for increased or decreased amplitude or power in the target band where required, namely rewarding increase or decrease with feedback depending on what the desired outcome might be (hurdling over the bar or limbo-ing under the bar in his helpful analogy). Usually two standard deviations in statistics is regarded as a target, but here alpha stands out, especially in the presence of elevated beta, where scores less than 1 in Z terms may be a target.  Adjusting Z score thresholds is designed to increase power, and there are helpful discussions of this with diagrams as usual and a case study of a memory issue.

Part II embarks as noted earlier with a discussion of using the amplified and filtered signal to match EEG signatures to common symptoms. Here, the two active electrodes collect the data, using the grounding earth lead as a reference point, usually attached to an ear or the mastoid, if the ear is not great signal-wise. Shared values are filtered out, and only the difference between the two reference leads or active leads, is accepted as data. As I noted again, the artifact issue has to be dealt with or the results are contaminated. Rejection of artifact can also be accomplished by using the common sEMG data and eliminating that as they share the signal and thus we don’t consider the common ground as meaningful, and this would apply to the bipolar montage. 

19 Channels share one common ground in qEEG, producing an encephalogram, covering epochs of 1 second, with a few epochs spreading across the record paper. I mentioned the concept of bins before, and these are filters that work via Hz readings, and can be single epoch measures or bandwidths of say 4-8Hz. FFT filters gather the whole recording and then divide, IIR, measuring first then dividing the frequency bins. Single Hz bins require the operator to find the bin with the red, high Z-score, and then create a range that encompasses that single Hz, and this is the training range, usually 3 Hz for accuracy purposes. Eyes closed measures of the posterior areas refer to the Posterior Dominant Rhythm, increasing with age from 6Hz or so up to 10Hz in the range from 1 year to 14 year ages. So at 11 Hz, the measure of 2.5 microvolts might be the highest of the 1-12Hz measures, and so constitutes the PDR. 

An extensive explanation of the common filtered bandwidths follows, starting with delta, then they others, and explaining some truths of the field: we don’t for instance train theta in a person with a history of seizure, or in certain areas such as the frontal lobes (making judgement areas sleepy, not a good look). Perhaps PTSD might prohibit theta training approaches at the risk of triggering unwanted imagery. SMR’s are also mentioned, as noted above, namely Sensorimotor Rhythm which may reflect a state of being internally orientated, another idling rhythm along with alpha, and are trained for anxiety, hyperactivity, phobias, seizures (idle not theta stim, so you can see the logic), pain, sleep issues and sensory integration, using C3-C4 electrodes.  Gamma training is one of the most interesting, requiring a cognitive load and rewarding 40Hz.

I noted filtered EEG components such as coherence, and the next chapter takes this on. Hypocoherence might be a sign of a learning issue if in the parietal areas, or poor memory in the temporal areas, or frontal lobes, perhaps OCD. Location is everything, and distance between homologous electrodes also matters, as T will be further apart than central electrodes for instance.

Having established such parameters of reasoning, the next chapter sets out to chart what EEG signatures might map to common symptoms and disorders.  Depression in the left hemisphere might require targeting T5 and P3 theta, if beyond 2sd’s, namely the Z-score threshold for abnormality appear on the absolute power score at those positions. LH theta is thus greater than RH theta. Prefrontal measures of theta might reflect poor motivation, poor executive functioning and so on.

Part III becomes more complex in technical terms, namely the editing of the raw EEG. While we may look for spindle complexes for instance in the raw EEG, namely normal issues, but artifact that reflects on sleepiness for instance is an issue, as are ECG or pulse artifacts, and spike and wave formation, and a host of other issues such as high amplitude mu waves. Editing thus refers to the removal of unwanted or confounding error variables so that the main effects of concern can be observed and dealt with in training. Examples are given in the next chapter, which is extensive.

Part IV deals with the more morphologically detailed aspects, namely the nervous system, brain structure and function, cortical and subcortical regions of interest for biofeedback, and brain networks of interest since this is a systemic intervention in terms of homeostasis and so on. Of most interest are the 10-20 placements and the correlating structures that underly these placements. So Pz would overlay the dorsal anterior cingulate, T5 and T6, the fusiform gyrus, and that would be over the inferior, posterior temporal lobes.

Part V refers to advanced training and protocol generation, now getting to the nub of it all.  Most importantly, the idea of treating the whole person is inserted here, vitally, so that the focus does not entirely land on the EEG data alone. The book ends with phase training.

As a how to manual, and abandoning the whole history lesson and other aspect of the first edition, this book succeeds brilliantly, owing to the way it is structured in part.  It is graduated from easy and simplistic to hard and complex, and then Part IV does this again, starting with basic science and expanding at a nice pace.

You should emerge, as a novice, from this book, as a person who is ready to begin training with a keen eye and knowledge for best practice. However, as Demos notes, you need to be trained, but here is the way to begin. The use of clear imagery is so necessary, and he doesn’t stint at all, with both colour and b&w images abounding. The complex index and addendums help tremendously, and the writing is clear, and flows so that understanding is gained in incrementally meaningful ways.

 

© 2019 Roy Sugarman

 

Roy Sugarman PhD, Clinical Neuropsychologist and Clinical Psychologist, Director: Applied Neuroscience, Performance Innovation Team, Team EXOS, USA