Commit b29efc
2026-04-13 23:45:13 Anonymous: Initial commit| /dev/null .. neurotech docs/neurotech platform design.md | |
| @@ 0,0 1,486 @@ | |
| + | # Neurotechnology Platform Design: From Markers to States |
| + | |
| + | ## Core Principle |
| + | |
| + | A neurotechnology platform should **not** treat individual neurofeedback markers as states. |
| + | |
| + | Instead, it should use the following pipeline: |
| + | |
| + | **marker / feature -> construct axis -> task-specific state** |
| + | |
| + | This is the core abstraction that turns a collection of EEG / fNIRS / physiological features into a reusable software foundation for neurofeedback, cognitive training, and clinical applications. |
| + | |
| + | --- |
| + | |
| + | ## The Problem This Solves |
| + | |
| + | Literature tables give us many useful markers: |
| + | |
| + | - SMR |
| + | - FMT |
| + | - Theta/Beta Ratio |
| + | - Upper Alpha |
| + | - FAA |
| + | - SCPs |
| + | - Alpha/Theta |
| + | - T3 Alpha / Temporal-Frontal Coherence |
| + | - HbO Up-Regulation |
| + | - Decoded EEG or fNIRS signals |
| + | - etc. |
| + | |
| + | But these are **not** directly the things an app should train. |
| + | |
| + | For example: |
| + | |
| + | - **SMR** is not the same thing as **calm focus** |
| + | - **FAA** is not the same thing as **good mood** |
| + | - **Theta/Beta Ratio** is not the same thing as **attention** |
| + | - **FMT** is not universally “up good” or “down good” |
| + | |
| + | These markers are better treated as: |
| + | |
| + | - **observations** |
| + | - **control handles** |
| + | - **evidence-backed proxies** |
| + | - **inputs to latent construct estimation** |
| + | |
| + | So the platform should infer **construct axes** and then map those axes into **task-specific states**. |
| + | |
| + | --- |
| + | |
| + | ## The Pipeline |
| + | |
| + | ## 1. Marker / Feature Layer |
| + | |
| + | This is the raw library of things the platform can compute. |
| + | |
| + | Examples: |
| + | - EEG bandpower |
| + | - individualized alpha |
| + | - theta/beta ratio |
| + | - SMR |
| + | - FAA |
| + | - coherence |
| + | - SCP amplitude |
| + | - HbO / HbR changes |
| + | - decoded multivariate patterns |
| + | - HRV |
| + | - respiration |
| + | - EDA |
| + | - movement load |
| + | - task performance metrics |
| + | |
| + | This layer answers: |
| + | - What can we measure? |
| + | - From where? |
| + | - With what latency? |
| + | - Under what task assumptions? |
| + | - For which populations or use cases? |
| + | - With what evidence strength? |
| + | |
| + | This should be implemented as a **feature registry**. |
| + | |
| + | --- |
| + | |
| + | ## 2. Construct Axis Layer |
| + | |
| + | Construct axes are reusable latent dimensions that markers contribute to. |
| + | |
| + | These axes are more stable and reusable than app-specific states. |
| + | |
| + | A list of potential shared axes: |
| + | |
| + | - **Arousal / activation** |
| + | - **Task engagement** |
| + | - **Cognitive control** |
| + | - **Calm focus / stable attentional readiness** |
| + | - **Motor automaticity** |
| + | - **Perceptual breadth / visuospatial readiness** |
| + | - **Affective regulation / emotional load** |
| + | - **Executive recruitment / cognitive load** |
| + | - **Fatigue / instability** |
| + | - **Sleep readiness / sleep stability** |
| + | - **Signal reliability** |
| + | |
| + | A single marker can contribute to multiple axes. |
| + | |
| + | Example: |
| + | - **SMR** may contribute to calm focus, motor stability, and sleep stability |
| + | - **FMT** may contribute to task engagement and cognitive control |
| + | - **FAA** may contribute to affective regulation |
| + | - **dlPFC HbO** may contribute to executive recruitment and cognitive load |
| + | |
| + | This is the key move: |
| + | **markers are grouped into latent constructs rather than treated as whole states** |
| + | |
| + | Similarly, each axis would be composed of multiple markers. |
| + | |
| + | For example, calm focus could potentially be derived from SMR, upper alpha and high-beta. |
| + | |
| + | --- |
| + | |
| + | ## 3. Task-Specific State Layer |
| + | |
| + | States are protocol-specific control abstractions built from combinations of axes. |
| + | |
| + | Examples: |
| + | |
| + | ### Athletic precision states |
| + | - under-engaged |
| + | - calm-focused |
| + | - over-aroused |
| + | - over-controlled |
| + | - automatic / in-the-pocket |
| + | - fatigued |
| + | |
| + | ### Cognitive enhancement / wellbeing states |
| + | - distracted |
| + | - stable attention |
| + | - mentally strained |
| + | - inward / meditative |
| + | - emotionally loaded |
| + | - sleep-ready |
| + | |
| + | ### Clinical states |
| + | - hyperaroused |
| + | - dysregulated |
| + | - avoidant / shut down |
| + | - regulated |
| + | - impulsive / distractible |
| + | - cognitively overloaded |
| + | |
| + | States are what the app or protocol actually responds to. |
| + | |
| + | --- |
| + | |
| + | ## Recommended Software Abstraction |
| + | |
| + | The platform should be organized as: |
| + | |
| + | **signals -> markers -> axes -> states -> feedback policy** |
| + | |
| + | ### Signals |
| + | Raw EEG / fNIRS / physiology / behavior |
| + | |
| + | ### Markers |
| + | Computed features and protocol-specific measurements |
| + | |
| + | ### Axes |
| + | Shared latent constructs estimated from marker combinations |
| + | |
| + | ### States |
| + | Task-specific regions or categories in axis space |
| + | |
| + | ### Feedback Policy |
| + | What the app does next: |
| + | - visual feedback |
| + | - audio feedback |
| + | - task adaptation |
| + | - protocol progression |
| + | - clinician recommendation |
| + | - stimulation policy |
| + | |
| + | --- |
| + | |
| + | ## Why This Matters |
| + | |
| + | This architecture prevents the platform from becoming: |
| + | |
| + | - a bag of unrelated protocol scripts |
| + | - a hardcoded “SMR app / FAA app / Theta-Beta app” zoo |
| + | - a brittle collection of narrow decoders |
| + | |
| + | Instead, it creates: |
| + | |
| + | - a shared ontology |
| + | - reusable estimators |
| + | - protocol portability |
| + | - room for personalization |
| + | - room for future learned representations or manifold models |
| + | |
| + | --- |
| + | |
| + | # Marker-to-Axis-to-State Ontology |
| + | |
| + | ## Canonical Axes |
| + | |
| + | | Canonical Axis | What It Represents | Example App-Level States | |
| + | |---|---|---| |
| + | | **Arousal / Activation** | How under-activated, optimal, or over-activated the person is | under-engaged, optimal, over-aroused | |
| + | | **Task Engagement** | Whether the person is actively on-task vs drifting or disengaging | distracted, engaged, mind-wandering | |
| + | | **Cognitive Control** | Degree of top-down task control and monitoring | unfocused, controlled, over-controlled | |
| + | | **Calm Focus / Stable Attentional Readiness** | Quiet, stable, low-noise task readiness | scattered, calm-focused, sedated | |
| + | | **Motor Automaticity** | Degree to which motor skill is flowing without verbal interference | effortful, automatic, choking | |
| + | | **Perceptual Breadth / Visuospatial Readiness** | Peripheral awareness, spatial monitoring, situational scanning | tunnel vision, broad awareness, diffuse attention | |
| + | | **Affective Regulation / Emotional Load** | Emotional distress, regulation success, approach vs avoidance | distressed, regulating, settled | |
| + | | **Executive Recruitment / Cognitive Load** | Working memory and prefrontal task recruitment | under-recruited, optimal, overloaded | |
| + | | **Fatigue / Instability** | Cognitive or performance drift, degradation, inconsistency | fresh, stable, fatigued | |
| + | | **Sleep Readiness / Sleep Stability** | Transition toward sleep and ability to maintain sleep-supportive physiology | alert, winding down, sleep-ready, restless | |
| + | | **Signal Reliability** | Whether the platform should trust the current signal estimate | usable, noisy, invalid | |
| + | |
| + | --- |
| + | |
| + | ## Internal Ontology Table |
| + | |
| + | | Marker | Domain | Canonical Axis | Typical Sign | Context Sensitivity | Confidence | Candidate Protocol States | |
| + | |---|---|---|---|---|---|---| |
| + | | **Sensorimotor Rhythm (SMR)** | Athletic, Wellbeing, Clinical | Calm Focus; Motor Automaticity; Sleep Stability | Usually positive for calm, stable, low-noise readiness | **Medium**. Can reflect useful stability, but may also look high in low-demand or passive states | **Medium-High** in sport; **Medium** in attention; **Low-Medium** in sleep | calm-focused, stable attention, automatic, sleep-stable, under-engaged if high without behavioral engagement | |
| + | | **Frontal Midline Theta (FMT)** | Athletic, Wellbeing | Task Engagement; Cognitive Control | **Bidirectional**. Can be beneficial when higher or lower depending on expertise and task | **High**. Must be calibrated to task and person | **Medium** | engaged, focused, over-controlled, flow-ready, cognitively strained | |
| + | | **Theta/Beta Ratio** | Athletic, Wellbeing, Clinical | Task Engagement; Distractibility / Attention Regulation | Lower ratio often better for focused alertness | **Medium**. Stronger in ADHD-like settings than general wellness | **Medium** | distractible, engaged, impulsive, task-ready | |
| + | | **Multi-Band Reaction Speed Profile**<br>(SMR and Beta1 up / Theta and Beta2 down) | Athletic | Calm Focus; Task Engagement; Fatigue / Instability | Positive when stable focused alertness increases and drift decreases | **Medium** | **Medium** | locked in, scattered, mentally cooked, reaction-ready | |
| + | | **Upper Alpha / Individualized Alpha** | Wellbeing, Athletic, Clinical | Calm Focus; Memory Readiness; Perceptual Stability | Usually positive for calm attentional control and working-memory support | **Medium**. Better when individualized around IAF | **Medium-High** | stable attention, deep work, broad readiness, calm cognitive control | |
| + | | **Alpha Band Up-Training / CVSA** | Athletic | Perceptual Breadth / Visuospatial Readiness | Positive for broad peripheral awareness and covert visuospatial attention | **Medium-High**. Likely task-specific and best with behavioral anchors | **Emerging** | tunnel vision, broad field awareness, scanning-ready | |
| + | | **COSMI Index** | Athletic | Calm Focus; Task Engagement; Motor Readiness | Positive when SMR rises and distracting/noisy bands drop | **Medium** | **Emerging** | reaction-ready, precise, stable, mentally noisy | |
| + | | **ACC Modulation / Arousal Regulation** | Athletic | Arousal / Activation | Positive when arousal moves toward an optimal band, not simply lower | **High**. The target is optimality, not one direction | **Emerging** | flat, optimal, overheated, panic-prone | |
| + | | **Left Temporal Alpha (T3)** | Athletic | Motor Automaticity | Usually positive for reducing verbal-analytic interference in precision skills | **High**. Most relevant in expert self-paced precision contexts | **Medium** | automatic, overthinking, choking, fluent execution | |
| + | | **Temporal-Frontal Coherence** | Athletic | Motor Automaticity; Cognitive Control | Often interpreted relative to reduced conscious overcontrol | **High** | **Medium** | automatic, effortful control, paralysis by analysis | |
| + | | **FAA (Frontal Alpha Asymmetry)** | Wellbeing, Clinical | Affective Regulation / Emotional Load | Protocol-dependent; often tied to healthier approach-oriented affect | **Medium-High**. Evidence is mixed and targetability is not perfectly settled | **Medium** | distressed, regulating, settled, pre-sleep emotionally downshifted | |
| + | | **Decoded EEG Emotion-State / Reappraisal Signal** | Wellbeing, Clinical | Affective Regulation; Cognitive Reappraisal | Positive when decoded affect-regulation pattern matches desired state | **High**. Depends on decoder training and calibration | **Emerging** | emotionally loaded, reappraising, resilient, settled | |
| + | | **Alpha/Theta Ratio / Alpha-Theta Training** | Wellbeing, Clinical | Arousal / Activation; Affective Regulation; Sleep Readiness | Often positive for downshifting, inwardness, relaxation | **Medium** | **Medium-Low to Medium** | relaxed inward, unwinding, meditative, sleep-ready | |
| + | | **SCPs (Slow Cortical Potentials)** | Wellbeing, Clinical, Accessibility / BCI | Task Engagement; Intentional Control | Depends on whether the protocol trains activation or deactivation | **Medium** | **Medium-High clinically** | engaged, release, intentional control, accessibility control state | |
| + | | **PCC / DMN Downregulation** | Wellbeing | Affective Regulation; Inward Attention; Mind-Wandering Control | Positive when self-referential drift decreases during meditation-like states | **High**. Measurement often indirect unless imaging is used | **Emerging** | mind-wandering, present, inward, meditative | |
| + | | **dlPFC HbO Up-Regulation** | Athletic, Wellbeing, Clinical | Executive Recruitment / Cognitive Load; Task Engagement | Positive up to a point. More is not always better if overload appears | **Medium** | **Medium / Emerging** | under-recruited, optimal executive engagement, overloaded | |
| + | | **Decoded Prefrontal fNIRS Patterns** | Wellbeing | Executive Recruitment; Cognitive Control | Positive when decoded control-related patterns strengthen | **High**. Heavily decoder-dependent | **Emerging** | anti-distraction, executive-ready, controlled, resilient | |
| + | | **Network-Based fNIRS Small-Worldness** | Wellbeing | Executive Recruitment; Cognitive Control; Fatigue / Instability | Positive when network efficiency aligns with lower cognitive load and stronger control | **High** | **Emerging** | efficient-control, overloaded, unstable-control | |
| + | | **SMR-Linked Sleep Stability / Spindle-Adjacent Training** | Wellbeing, Clinical | Sleep Readiness / Sleep Stability | Potentially positive for sleep-supportive stability, but evidence is mixed | **Medium** | **Mixed** | winding down, sleep-stable, restless | |
| + | | **Autism-Oriented EEG Self-Regulation Targets**<br>(SCP, beta/theta, mu / alpha etc.) | Clinical | Task Engagement; Affective Regulation; Executive Recruitment | Protocol-dependent | **High**. Highly population- and target-specific | **Emerging-Medium** | engaged, dysregulated, more regulated, overloaded | |
| + | | **ADHD-Oriented EEG Targets**<br>(Theta/Beta, SMR, SCPs) | Clinical | Task Engagement; Cognitive Control; Fatigue / Instability | Protocol-dependent; usually lower distractibility and higher control are desirable | **Medium** | **Mixed / Low** | distractible, task-engaged, impulsive, fatigued | |
| + | | **Tinnitus Alpha/Delta Targets** | Clinical | Affective Regulation; Sensory Distress / Symptom Load | Protocol-dependent and likely indirect | **High** | **Emerging / Mixed** | distressed, symptom-loaded, more regulated | |
| + | | **MCI Cognitive EEG Targets**<br>(alpha, beta, SMR/theta combinations) | Clinical | Executive Recruitment; Memory Readiness; Task Engagement | Usually positive when supporting cognitive engagement and memory stability | **Medium** | **Emerging-Medium** | under-recruited, cognitively engaged, fatigued | |
| + | |
| + | --- |
| + | |
| + | # How to Use This in Practice |
| + | |
| + | ## Rule 1: Do Not Build Apps Around Single Markers |
| + | |
| + | Do **not** build: |
| + | - an SMR app |
| + | - an FAA app |
| + | - a theta/beta app |
| + | |
| + | Build: |
| + | - a calm-focus estimator |
| + | - an overcontrol detector |
| + | - a sleep-readiness estimator |
| + | - an affective regulation estimator |
| + | - a competition-readiness estimator |
| + | |
| + | Markers should sit underneath the estimator. |
| + | |
| + | --- |
| + | |
| + | ## Rule 2: Build Shared Axes First |
| + | |
| + | A platform should estimate reusable axes before it estimates task states. |
| + | |
| + | Recommended shared axes for version 1: |
| + | |
| + | - Arousal / Activation |
| + | - Task Engagement |
| + | - Cognitive Control |
| + | - Calm Focus |
| + | - Motor Automaticity |
| + | - Affective Regulation |
| + | - Executive Recruitment |
| + | - Fatigue / Instability |
| + | - Sleep Readiness |
| + | - Signal Reliability |
| + | |
| + | These axes can then be reused across: |
| + | - sport |
| + | - wellbeing |
| + | - cognitive enhancement |
| + | - clinical training |
| + | |
| + | --- |
| + | |
| + | ## Rule 3: States Are Domain-Specific |
| + | |
| + | ### Example sport states |
| + | - under-engaged |
| + | - calm-focused |
| + | - over-aroused |
| + | - over-controlled |
| + | - automatic |
| + | - fatigued |
| + | |
| + | ### Example wellbeing states |
| + | - distracted |
| + | - stable attention |
| + | - mentally strained |
| + | - inward / meditative |
| + | - emotionally loaded |
| + | - sleep-ready |
| + | |
| + | ### Example clinical states |
| + | - hyperaroused |
| + | - dysregulated |
| + | - avoidant / shut down |
| + | - regulated |
| + | - impulsive / distractible |
| + | - cognitively overloaded |
| + | |
| + | --- |
| + | |
| + | # Worked Example: Athletic Performance |
| + | |
| + | ## Goal State |
| + | **Calm-focused and competition-ready** |
| + | |
| + | This should not be inferred from SMR alone. |
| + | |
| + | It should be estimated from multiple axes, for example: |
| + | |
| + | - **Calm Focus** |
| + | - **Motor Automaticity** |
| + | - **Arousal Optimality** |
| + | - **Fatigue / Instability** |
| + | |
| + | ## Example Marker Contributions |
| + | |
| + | ### Calm Focus |
| + | - SMR |
| + | - Upper Alpha |
| + | - Theta/Beta Ratio |
| + | - COSMI components |
| + | |
| + | ### Motor Automaticity |
| + | - T3 Alpha |
| + | - Temporal-Frontal Coherence |
| + | - Expert-specific FMT effects |
| + | |
| + | ### Arousal Optimality |
| + | - ACC modulation |
| + | - stress-sensitive beta features |
| + | - optional HRV / respiration if available |
| + | |
| + | ### Fatigue / Instability |
| + | - performance drift |
| + | - RT variability |
| + | - dlPFC HbO load |
| + | - multi-band degradation over time |
| + | |
| + | ## Example State Definition |
| + | |
| + | A rough internal rule might be: |
| + | |
| + | **competition-ready** when: |
| + | - Calm Focus is high |
| + | - Automaticity is high |
| + | - Arousal is within an optimal band |
| + | - Fatigue is low |
| + | - Signal quality is acceptable |
| + | |
| + | This means: |
| + | |
| + | **SMR does not directly become “competition-ready”** |
| + | It becomes one weighted contributor to the **Calm Focus** axis, which then contributes to the final **competition-ready** state. |
| + | |
| + | --- |
| + | |
| + | # Worked Example: From SMR to Calm Focus |
| + | |
| + | ## Wrong approach |
| + | **SMR = calm focus** |
| + | |
| + | ## Better approach |
| + | SMR contributes to a **Calm Focus** construct together with other markers. |
| + | |
| + | Example conceptual composite: |
| + | |
| + | **Calm Focus** is estimated from: |
| + | - SMR |
| + | - Upper Alpha |
| + | - High-Beta or tension marker |
| + | - Theta/Beta Ratio |
| + | - optional EMG / motion noise |
| + | - optional behavioral stability |
| + | |
| + | Then the athlete’s state is determined from that axis plus others. |
| + | |
| + | This gives: |
| + | |
| + | **SMR -> Calm Focus axis -> Competition-Ready state** |
| + | |
| + | But in practice the real flow is: |
| + | |
| + | **SMR + other markers -> Calm Focus axis** |
| + | then |
| + | **Calm Focus + other axes -> Competition-Ready state** |
| + | |
| + | --- |
| + | |
| + | # Platform Design Recommendation |
| + | |
| + | The platform should implement: |
| + | |
| + | ## 1. Marker Registry |
| + | Stores: |
| + | - what the marker is |
| + | - how it is computed |
| + | - where it is measured |
| + | - required preprocessing |
| + | - domain relevance |
| + | - evidence confidence |
| + | - known caveats |
| + | |
| + | ## 2. Axis Estimators |
| + | Reusable models that combine markers into construct axes. |
| + | |
| + | Examples: |
| + | - Calm Focus Estimator |
| + | - Arousal Estimator |
| + | - Affective Regulation Estimator |
| + | - Executive Recruitment Estimator |
| + | |
| + | ## 3. State Decoders |
| + | Protocol-specific mappings from axes to app states. |
| + | |
| + | Examples: |
| + | - Sport readiness decoder |
| + | - Sleep readiness decoder |
| + | - Mood regulation decoder |
| + | - ADHD attention-state decoder |
| + | - Trauma regulation decoder |
| + | |
| + | ## 4. Feedback Policy Engine |
| + | Decides what the app does once state is inferred. |
| + | |
| + | Examples: |
| + | - increase / decrease feedback intensity |
| + | - adapt task difficulty |
| + | - hold state |
| + | - encourage down-regulation |
| + | - reinforce best-state similarity |
| + | - trigger clinician note or safety pause |
| + | |
| + | --- |
| + | |
| + | # Summary |
| + | |
| + | ## Markers |
| + | Evidence-backed measurements and control handles |
| + | |
| + | ## Construct Axes |
| + | Reusable latent dimensions that markers contribute to |
| + | |
| + | ## Task-Specific States |
| + | Protocol-level control abstractions built from axes |
| + | |
| + | ## Final Rule |
| + | |
| + | **Markers are not states.** |
| + | **Axes are reusable constructs.** |
| + | **States are protocol-specific regions in axis space.** |
| + | |
| + | So the correct abstraction is: |
| + | |
| + | **signals -> markers -> axes -> states -> feedback policy** |
| + | |
| + | That is the foundation for a neurotechnology platform that supports: |
| + | - athletic performance products |
| + | - cognitive enhancement products |
| + | - wellbeing products |
| + | - clinical neurofeedback products |
| + | - future manifold / learned-representation layers |