BCIH: "Introduction (BCI Basics)" [Chp. 1 Sec. 1]

[This post based on the Brain-Computer Interfaces Handbook, Section 1.1]

Why are Brain-Computer Interfaces (BCIs) important, and how do we classify them? 

BCIs enable humans to affect the world in ways other than physical movements by providing new output pathways. These new 'output pathways' can enable us to restore lost functions to patients with neural damage, as well as provide new capabilities to healthy users. One example of how these new output pathways can be used is this 2015 study in which a BCI was used to control a quadcopter. An analysis of that BCI based on the classification criteria presented here is at the end of this post.

Usually, BCIs use the following pipeline to read and understand data:

In this post, we'll analyze how we classify BCIs based on the first two steps of that pipeline.

Step 1

For the first step (Brain Activity Pattern Generation), there are two main classes of how the BCI affects the brain activity generation step:

1. Active/re-active BCIs. In active BCIs, the user actively tries to generate activity in a specific way to control an application. In re-active BCIs, the BCI provides stimulus to the user and then reads the subsequently generated activity.
2. Passive BCIs. In passive BCIs, brain activity serves is just an additional input source for an application, and no efforts are made to elicit specific activity from the user.

Step 2

For the second step (Signal Acquisition), the author distinguishes BCI signal acquisition methods using five criteria:

1. Brain Signal Pattern
2. Stimulus Modality
3. Mode of Operation
4. Operation Strategy
5. Recording Method

The rest of this post will focus on explaining the subcategories of each criterion.

Brain Signal Pattern

When reading brain signals, one can focus on different activity patterns, each of which typically occurs in a specific area of the brain. The author mentions four activity pattern types:

1. P300 Event-Related Potentials (ERPs)
1. Signals that reach a positive peak 300ms after stimulus 
2. Steady-State Evoked Potentials (SSEPs)
1. Occurs in response to a specific, external sensory input (as opposed to a spontaneous one). Three types:
1. Steady-State Auditory Evoked Potentials (SSAEPs)
1. Originate in brainstem, respond to sound input
2. Steady-State Visually Evoked Potentials (SSVEPs)
1. Originate in occipital lobe, responds to regularly repeating visual input (author's example: a flashing light)
3. Steady-State Somatosensory Evoked Potentials (SSSEPs)
1. Originates (typically) in primary somatosensory cortex (S1), responds to tactile input (best given on glabrous skin, i.e., skin free from hair)
3. Slow Cortical Potentials (SCPs)
1. Slow (300ms to several second) potentials in response to a stimulus that is either a real, physical stimuli ('exogenous') or an expectation of a stimuli ('endogenous'). Since the latter can be consciously generated, one can train to produce this type of signal pattern on demand. Originates in upper cortical layers.
4. Sensorimotor Rhythm (SMR)
1. Oscillations that originate in sensorimotor cortex

Stimulus Modality

For active and re-active BCIs, stimulus can be presented to the user via visual, auditory, or tactile modalities, or some combination of the above (a hybrid modality). 

Mode of Operation

A BCI's mode of operation is the 'underlying way in which brain signals are elicited'.

1. A synchronous BCI presents stimulus and then reads the resulting activity ('cue-based').
2. An asynchronous BCI, on the other hand, does not provide stimuli, instead allowing the user to guide the interaction with conscious changes in their thought patterns ('self-paced'). 

Operation Strategy

Operation strategy is how the BCI prompts signals from the user. There are three types:

1. Selective Attention: A constant stimuli is presented, and the user focuses their attention on it ('selects' it).
2. Cognitive Efforts: The BCI presents biofeedback to help the user train themselves to maintain a certain activity pattern.
3. Motor Imagery: The user is asked to imagine doing certain motor movements to generate readable motor cortex activity.

Recording Method

After deciding what type of activity to read and how to generate it, one must decide the optimal way to physically read the data from the brain. The two recording options are:

1. Invasive, in which surgical work is needed to install the data reading hardware (e.g., ECoG or optical neuronal recording).
2. Non-invasive, in which no surgical work is required to read brain data. Some examples are those that attach to the surface of the body (EEG or fNIRS), use magnets or magnetic fields (MEG and fMRI), or use injected radioactive tracers and a detector (PET).

Summary

Brain-Computer Interfaces can solicit and gather data in a variety of different ways. Data can be solicited either by providing explicit stimuli to the user, or by reading data from a user's ongoing interaction with another system (no explicit stimulus). Data can be read from a variety of brain regions and using several different techniques, which fall different places on a continuum of invasiveness to the user. 

Going back to the 2015 quadcopter study mentioned in the introduction, we can now apply classification to the first two stages of this quadcopter-controlling BCI (this BCI is multi-modal, so encompasses several of the above elements): 

1. Brain Activity Pattern Generation: Active (Motor Imagery, Physical Blinking), Re-Active (SSVEP With 12.4Hz and 18Hz Lights)
2. Signal Acquisition:
1. Brain Signal Pattern: 
1. SSVEP (used to switch between quadcopter rising/falling modes).
2. Eye-Blinking Activity (read from frontal/prefrontal lobes, switches between the rising/falling and turning modes).
3. Sensorimotor Rhythm (motor imagery signals, which control turning left or right).
2. Stimulus Modality: Visual (LED lights, quadcopter's movements).
3. Mode of Operation: Asynchronous.
4. Operation Strategy: Selective Attention (LED lights), Motor Imagery (Left/right switch).
5. Recording Method: Non-invasive.

As we can see from this example, this classification system is useful in breaking down a BCI into its individual parts. Such a breakdown enables an abstraction of BCIs into a form that is easy to understand quickly and use for direct comparisons between BCI approaches.

Sources

1. Brain Computer Interfaces Handbook, Chapter 1
2. Slow Cortical Potentials: http://biofeedbackfederationofeurope.blogspot.com/2014/04/introduction-to-slow-cortical.html
3. Quadcopter: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6554428/
4. Eye Blinking and EEG: http://www.iaeng.org/publication/WCE2010/WCE2010_pp63-66.pdf

Acronyms

1. ECoG: Electrocochleography
2. EEG: Electroencephalogram
3. fNIRS: Near-infrared spectroscopy
4. fMRI: Functional magnetic resonance imaging
5. MEG: Magnetoencephalography
6. PET: Positron emission tomography