Make the paradigms more broadly

[bruAristimunha]

At the moment, we have motor imagery, P300, ssvep, cvep and rest stage.

We need to make this more flexible; motor imagery should be under the umbrella imagery, and P300 should be ERP.

Perhaps ssvep should also be more general, but I need to think about it more.

This is necessary to accommodate new data categories, such as m3cv, speech imagery, more competition dataset, and perhaps something more general.

[PierreGtch]

We can do a split between self-paced, stimulation-based and passive BCIs.

@thijor would argue that SSVEP is frequency tagging, ERP is temporal tagging and cVEP is time-frequency tagging.

@bruAristimunha, do you want to want to restructure the implementation the different paradigms? maybe we could create a tree structure for the different paradigms.
At the moment we have a mini tree structure (motor imagery -> left-right MI) but the paradigm names don't reflect that.

[bruAristimunha]

I made a small modification for #1057, I introduced some small aliases called imagery, and I created the speech imagery as requested by someone interested in moabb.

I would be super happy to re-design the paradigm if we can think together about the "neural-signature" related to each paradigm.... I know that Fabien Lotte has a lot of theory about the motor imagery and ERD/ERS...

[PierreGtch]

Here is a first draft. What do you think?

BCI paradigms
│
├── Self-paced (endogenous, user-initiated)
│   ├── Sensorimotor
│   │   ├── Motor Imagery (MI)
│   │   │   ├── Kinesthetic (imagining the feeling of doing the movement from a first person)
│   │   │   ├── Visual MI (imagining seeing the movement)
│   │   │   └── Fine-grained MI (individual fingers, grasp types)
│   │   ├── Motor execution / attempted movement
│   │   │   └── Movement-Related Cortical Potentials (MRCP, BP/CNV)
│   │   └── Sensorimotor rhythm (SMR) self-regulation / ERD-ERS
│   ├── Cognitive / mental tasks
│   │   ├── Mental arithmetic
│   │   ├── Mental rotation / spatial navigation imagery
│   │   ├── Word generation / verbal fluency
│   │   ├── Music or melody imagery
│   │   └── Imagined/inner/covert speech
│   └── Slow Cortical Potential (SCP) self-regulation  (Birbaumer's TTD)
│
├── Stimulation-based (exogenous, evoked)
│   ├── Transient evoked potentials (tVEP, AEP, tactile EP)
│   ├── Oddball / P300 (row-column spellers, region-based, RSVP)
│   ├── Steady-state responses (SSVEP / ASSR / SSSEP — frequency tagging)
│   ├── Code-modulated responses (c-VEP, random-VEP)
│   ├── Motion-onset responses (mVEP)
│   └── Continuous stimulus tracking (Auditory Attention Decoding, envelope/TRF)
│
└── Passive (implicit monitoring, no intentional control)
    ├── Workload / cognitive load
    ├── Vigilance / drowsiness / fatigue
    ├── Attention & engagement (incl. flow)
    ├── Affect / emotion recognition
    ├── Stress
    ├── Sleep staging
    ├── Error-related potentials (ErrP / feedback ErrN)
    └── Memory / encoding state (subsequent-memory effect)

Do not fit a simple tree but must be represented in MOABB:

• Modality is orthogonal. Any stimulation-based paradigm can in principle be driven through vision, audition, or touch (e.g., P300 has visual, auditory, and tactile variants; steady-state has SSVEP/ASSR/SSSEP).
• Hybrid BCIs combine paradigms from different branches — e.g., MI + SSVEP for extra commands, or an active BCI gated by a passive workload monitor.

Caveats:

• ErrP is technically stimulus-evoked but is almost always deployed passively as a correction signal on top of another BCI, so it lives under passive here.
• Imagined speech could sit under sensorimotor on a motor-theory reading, but is more commonly grouped with cognitive tasks.
• Neurofeedback is a training framework applied on top of SMR/SCP rather than its own paradigm.

[bruAristimunha]

It is okay proposal, but I think I would prefer to stay nearby some of the ongoing ontology's to avoid had to rethink about this on future.

I know that there is some categorization on HED, there is the definition that will be release by Jared, an there is this research group that it is pushing some good categorization in some biggg project on the eeg 101.

[bruAristimunha]

https://www.eeg101.eu/

[bruAristimunha]

I will send the email with you in copy!

[jarod1212]

Hi! Here’s a proposal based on how we usually categorize tasks in the team.

BCI
│
├── Synchronous
│ ├── Exogenous
│ │ ├── Oddball / P300 (Row-column spellers, RSVP)
│ │ ├── Transient evoked potentials (tVEP, AEP)
│ │ ├── Steady-state responses (SSVEP / ASSR / frequency tagging)
│ │ └── Code-modulated responses (c-VEP, random-VEP)
│ │
│ └── Cued Endogenous Tasks
│ ├── Cued Motor Imagery (e.g., classic Graz BCI "arrow" paradigms)
│ └── Cued Cognitive Tasks (Mental math, spatial navigation done in strict trials)
│
├── Asynchronous
│ ├── The "Brain Switch" / Event Detection
│ │ ├── Movement-Related Cortical Potentials (MRCP, BP/CNV)
│ │ └── Asynchronous Motor Imagery
│ │
│ └── Continuous State Modulation
│ ├── Sensorimotor rhythm (SMR) self-regulation / ERD-ERS
│ ├── Slow Cortical Potential (SCP) self-regulation
│ └── Sustained inner speech / covert speech
│
└── Psychophysiological Monitoring
├── System Interaction Assessment
│ └── Error-related potentials (ErrP)
│
└── Passive BCI
├── Workload / cognitive load
├── Vigilance / drowsiness / fatigue
├── Attention & engagement (incl. flow)
├── Affect / emotion recognition
├── Stress
├── Sleep staging
└── Memory / encoding state

[bruAristimunha]

many thanks @jarod1212 ;)

[PierreGtch]

I think the "cued" vs. "self=paced" distinction is important to make indeed!

[bruAristimunha]

https://www.hedtags.org/hed-task/criteria/task_criteria.html#the-inclusion-test