BCI Brain Signal Targets: Part I

The field of BCI has made impressive and significant progress in artificial movement, communication, and other practical applications. Nonetheless, effective translation of these interfaces from proof-of concept prototypes into reliable applications remains elusive (Chavarriaga et. al., 2016).

What is the cause of this underperformance? Some in the field have argued the user’s role in BCI control has been overlooked, and that taking into account user’s mental states and skills could have a substantial impact in improving BCI efficiency, effectiveness and usability (Lotte, Jeunet, Mladenovic et al., 2018).

I would extent this point further. The fundamental reason for the shortcoming in BCI performance, I argue, is that the mind (including that of a BCI user) is poorly understood (Poldrack & Yarkoni, 2016). Because of this, even a dramatic turn of focus toward the user’s mental state will have limited effect. To take full advantage of the user’s mind, one must first be able to define it — clearly, comprehensively, and accurately.

Definitions of the mind can be used as (mind) classifiers, labels, or categories. These I argue are central to BCI development. Defining user state of mind “signatures,” during tasks (such as “reach arm forward,” or “imagine the letter g”), and using these to classify the brain signal, is the key to improving BCI performance. Furthermore, I argue the mind can be defined — with accuracy, and precision. Therefore, state of mind classifiers can (a) be developed, (b) used to classify a given brain signal, and (c) also be “targeted” in advance (by the BCI designer, trainer, and user).

Typical user states of mind (perception, emotion, imagination, motivation, etc.) can be used as BCI brain signal classifier categories — because the brain signal directly reflects the user’s mind. State of mind classifier (a range of perception/emotion/imagination etc.) = a corresponding brain signal range.

Mind/brain signal classifiers, or “targets,” can be designed that are easy-to-hit (strongly and selectively), and that can be adapted flexibly to match the person, environment, situation, activity, task, device capabilities, and other variables.

I would further argue mind/brain signal targets can be defined for any BCI task. Such targets can be built into device design, training, and user control during device operation.

It’s important to note the task the user is engaged in is only one aspect of mind. State of mind also includes perception, emotional state, goals, attention, executive control, imagination, intention and many other components. Brain signal reflects the user’s mind much more so than the task the mind is engaged in. Optimal brain signal targets are of mind not task.

Although the mind/brain signal connection may seem obvious, it is a underappreciated issue within most of the BCI community. The mind is seldom recognized as having anything to do with, let alone be closely connected to, the brain signal. This is a huge problem, because the brain signal in fact mirrors a person’s mental state. This is basic cognitive neuroscience; yet this simple fact is seldom voiced clearly due to the materialist paradigm. The physical brain is seen as all the mind is. The conscious mind either doesn’t exist at all, or is trivial (perhaps a hallucination or illusion).

Not only do on-target states of mind (ex: feeling calm, focused, motivated…) take a back seat, so do off-target states of mind (ex: feeling frustrated, impatient, fatigued…). Off-target states can be a large part of the mind during a task. Yet they are seen, inevitably, as (vaguely-defined) noise. This is a wasted opportunity, because these off-target signals — if defined — can then be categorized as either noise, or as (part of) the signal. But this depends on their being recognized and defined as aspects of the user’s mind in the first place.

The way forward to developing more optimal — and useful — brain signal targets is to place the user’s mind at the center of the process. Targets based on the mind will not only enhance brain signal classification; but they allow the brain signal to be more consciously and skillfully used: by BCI designer, teacher, and user alike.

Given state of mind targets are reflected by corresponding brain signal targets, the question is how to optimally define the mind. This requires taking a step back to look at the mind: what is it, and how does it connects to the brain?

The mind does in fact exist as a phenomenon. Yet strangely, as all of neuroscience research indicates, it’s contents & function are located within the physical brain. Therefore, any subjective mental state or process would have to be mirrored by some aspect of the brain.

Furthermore, given the mind is expressed inside the brain, it has to have a neural correlate. For example, consider an intention — to “move my hand left forward.” This would need a brain signal that corresponds to, or represents, it. Otherwise, that particular intention could not cause that same movement, via the motor cortex and spine. Intentioin creates movement from WITHIN the brain. If an intention were manifest anywhere else but besides the brain it would not be able to generate that particular movement.

Confusion arises when brain and mind, conceptually, are muddled together. Brain is seen as “the mind.” This makes some sense in that, from a third person perspective, the mind is nowhere to be found. It can only be inferred. There IS no perception, thought, feeling, intention or any other aspect of mind to be seen from a third person view. There are no mental processes lurking among the neurons, or within the neural substrate.

Because of the dominance of the third-person objective view, and because the mind lacks a neural correlate, it is seen alternately as a non-entity, an inconvenient fact of the brain, and a nuisance to be ignored when possible. To understand the brain, the prevailing view is to ignore the mind and look to the brain; hopefully once enough brain data is gathered its relationship to the mind will become clear.

However, what if the subjective mind, and its neural correlate, are two sides of the same coin? What would this mean for the idea of brain signal “targets”? In principle, it means any aspect of the mind is — simultaneously — an aspect of brain activity, and corresponding brain signal. A given intention, emotion etc. = a given brain signal. Thus, when a BCI user is instructed to “intend to move your left arm forward,” she simultaneously generates a corresponding brain signal. Brain signal targets are actually mind/brain signal targets.

What is the most likely candidate for the neural correlate of the mind? It’s known functional neural networks — regions of large scale neural activity acting in synchrony and coordination — play a dominant role in real-time brain function. Therefore it is safe to assume FNNs and the mind are closely connected, if not one and the same phenomenon.

As a particular intention/FNN is repeated over time, it forms a group or range of similar intentions/FNN activity. A “move my left hand” intention would come to represent a mind “signature” — and a corresponding FNN range signature.

In short, the mind (ex: a movement imagination/intention & accompanying perception, emotion etc.) = a set of (active) functional neural network ranges = a corresponding set of brain signal ranges = a potential BCI brain signal target.

To the extent the user’s mind (not simply task) is recognized as the target, and defined accurately, brain signal “targets” for the BCI designer, trainer, and user to shoot for will be more useful and accurate as well. The good news is, mind/brain signal targets can be defined, with accuracy and precision.

References

Chavarriaga, R., Fried, O., Kleih, S., Lotte, F., Scherer, R. (2016). Heading for new shores! Overcoming pitfalls in bci design. Brain-Computer Interfaces, 4, 60.

Lotte, F., Jeunet, C., Mladenovic, J., N’Kaoua, B., Pillette, L. (2018). A BCI challenge for the signal processing community: considering the user in the loop. Signal Processing and Machine Learning for Brain-Machine Interfaces, IET pp. 1-2

Poldrack, R.A., Yarkoni, T. (2016). From brain maps to cognitive ontologies: informatics and the search for mental structure. Annual Review of Psychology, 67, 587.

Leave a Reply

Your email address will not be published. Required fields are marked *