The Evolution of Protocol-based Training

by Siegfried Othmer | June 21st, 2006

The pace of recent developments in terms of clinical approaches calls for an occasional respite to find one’s bearings and recover perspective. First of all, I am amazed that the pace of new developments in this field is not slowing down, and secondly I am surprised that the pace of development of our own approach of mechanisms-based or protocol-based training is not plateauing either.

Looking back there have been several key milestones in the development of our own approach, and each of these has contributed in comparable measure to our clinical efficacy. The first departure to which we contributed some measure of authorship was the natural pairing of “C3beta” with “C4SMR,” using differential reinforcement parameters. The second was the particularization of our work with ADHD with bipolar placements that took us off the central strip, namely the combination of C3-Fz with C4-Pz. The third was the discovery of the pairing of T3-Fp1 with T4-Fp1 for the large category we labeled the instabilities. The fourth was the discovery of the importance of reward frequency optimization in the extension of our work to treatment-refractory populations. The fifth was the discovery of the value of inter-hemispheric placement for addressing our most intractable instabilities.

Each of these stages of progress also effected a change in the operative model, which in turn brought about a refurbishment of the already established procedures. For example, frequency optimization was essential to making inter-hemispheric training as useful as it has turned out to be. Once that was in place, however, the new findings also had obvious implications for how lateralized training was to be optimized as well. It did not take long to confirm that the lateralized placements optimized at frequencies that were predictable on the basis on the inter-hemispheric trainings. Now each of these can find their natural place in the sun, and perhaps unsurprisingly we find ourselves refurbishing the old placements, albeit with new insights and with more deliberate optimization. So the new “standard” pairing of “left-frontal” with “right-parietal” training turns out to be T3-Fp1 with T4-P4. The pedigree is apparent, but one might still think of this as the sixth new departure.

The current excitement is around the fact that the range of frequencies of 0-3 Hz contains its own world of mysteries, and may be the key to many of those who still challenge us in the clinic. It has long been clear to us that the brain organizes itself in a much more fine-grained manner in the frequency domain than we are able to determine with our techniques. We are unfortunately up against our own “uncertainty principle” when it comes to the analysis of data in the time-frequency domain. When we turn up the gain on frequency resolution we lose time resolution, and vice versa. So we can always only approximate what is actually going on with EEGs that fluctuate rapidly in both the frequency and time domains. We do get hints of what is actually there, and these hints persuade us that the brain is accustomed to organizing its affairs down to better than the 0.1-Hz level. That in turn means that we can consider the 0-3 Hz bandpass to consist of many subsidiary bands, and each of these should be subject to operant conditioning.

Some of the above developmental stages were contingent on evolution in hardware and software. At the outset [1985] we did not build flexibility into our reward bands simply because we did not think we would ever need it. On the hardware side, we did not provide for training at very low EEG frequencies early on because it did not occur to us that anyone would ever want to do that. Fortunately, the NeuroAmp was designed with low-frequency training as one requirement to be met (it is actually dc-coupled), and it turned out to be just in time to meet our emerging needs as we are now interested in training down to 0.3 Hz.

All of the above discussion has been concerned with reward-based feedback. On the inhibit-side there have been two significant developments: 1) the adoption of wideband inhibits, and 2) the more recent adoption of multiple inhibit bands, following Val Brown. Finally, the developments on the reward and inhibit side come together in a simultaneous two-channel training strategy such as was first adopted by Chuck Davis. By training on the difference signal between two channels we carry forward the reward schema we have been implementing with single-channel bipolar placement. By inhibiting on both the sum and the difference channels we “contain” the excursions in EEG irrespective of relative phase between the channels.

The convergence of these strategies in two-channel training on the central strip should retain all of the benefits of our inter-hemispheric single-channel work, and it should give us much of the benefit of Val Brown’s approach with a combination of multiple inhibit bands with dynamic thresholding. By balancing our strong reward-based training with optimized inhibit-based training we ease the burden of decision-making for the clinician. In a quandary, the clinician can always back off to an inhibit-only strategy where the software bears the entire burden. In this manner, we can ease the entry of novice clinicians even as the overall training strategy bulks up in complexity beyond what we had anticipated.

We remain in a continuing debate on the issue of clinical complexity. On the one hand, we want to be able to respond to any clinical challenge that crosses our threshold. On the other, we have to give continuing recognition to the observation that most of our training objectives are in fact met with the basic techniques. This should be even more the case once a more proper balance is restored between inhibit and reward-based training with the transition to two-channel training. That transition had been modeled for us not only by Chuck Davis but then by Val Brown as well. It was only the combination of power and simplicity of the reward-based training that kept us from transitioning sooner to the two-channel work. It has been waiting to happen for a long time.

With a near-doubling of the clinical complexity in going from three to five or six electrodes on the head, one hopes for more than a doubling of overall clinical effectiveness. That’s a tall order, and may in fact be beyond reach. But there will be a niche where this capability will be advantageous, and we’ll just have to tease that out. The most obvious area of application is where T3-T4 has been at its best, namely the most treatment-resistant instabilities. The conversation continues.

Leave a Reply