Visual Snow Research Portal
Multi-Dimensional Markov Cluster (MCL) — 3D to 10D State Spaces
Patient trajectories modelled as stochastic Markov chains in n-dimensional symptom vector space (3D–10D). Each node is a symptom cluster; edges carry weekly transition probabilities.
Load demo data or sync from the clinical portal to run clustering.
State Transition Matrix
Cluster Group Definitions
Run MCL to see cluster definitions.
3D Trajectory Plot
Patient state paths through symptom space — each line is one patient's weekly trajectory.
Run MCL to render trajectories.
Steady-State Distribution
Long-run probability of occupying each cluster state — proxy for population-level outcome forecast.
Convergence & Aversion Detection
Identifies patients on divergent (worsening) vs. convergent (improving) trajectories using eigenvalue decomposition of the transition matrix.
Run MCL first.
Silhouette k-Selection
Mean silhouette coefficient (Rousseeuw 1987) for k = 2–6. Higher score = better cluster separation. Highlighted bar = recommended k. Recomputed after each MCL run.
Monte Carlo AR Filter Optimisation
Stochastic simulation of phase-cancelling visual field generation for AR passthrough. Each trial samples noise parameters, renders a candidate filter set, and scores it against patient-specific symptom vectors. A feedback loop selects the top-performing filter configurations for iterative refinement.
Score Distribution
Feedback Loop Convergence
Top Filter Configurations
Best-performing AR negation filter parameter sets, ranked by score. Each row defines a candidate filter for hardware deployment.
Run simulation to populate filter candidates.
Efficiency Map
2D efficiency surface across the two most important filter parameters. Bright = highest predicted suppression score.
Feedback Loop Summary
Run simulation first.
Homunculus AI — Digital Twin Experiment Environment
In-silico patient agent driven by cohort-derived behavioural models. Tweak lifestyle and environmental parameters and observe predicted changes in VSS symptom trajectory before any human trial.
Behavioural Parameter Controls
Predicted Symptom Trajectory
Behavioural Pattern Delta Report
Change relative to baseline patient profile.
Run simulation first.
Exposure Risk Heatmap
Predicted VSS flare risk by day-of-week and hour-of-day given current parameter set.
VSS Symptom Progression Window
Live view of snow intensity and selected co-occurring symptoms across time. Use this to identify phase transitions, exacerbation windows, and remission periods.
Snow Intensity Timeline
Primary VSS metric. Dashed threshold lines mark mild / moderate / severe clinical zones.
Key Symptom Co-Progression
Select symptoms overlaid on the snow intensity baseline. Identifies temporal coupling and lag relationships.
Phase Transition Detector
Identifies statistically significant step changes in snow intensity using sequential t-test. Red = exacerbation onset, teal = remission onset.
Render window first.
Remission Probability Curve
Kaplan-Meier-style estimate of remaining below the mild threshold. Based on cohort transition rates.
Panic Attack Detection — AR + Biomarker Fusion
Multi-modal panic prediction framework fusing wearable haemodynamic data, AR spatial tracking, and mobile behavioural signals. Simulates the sensor fusion pipeline described in Részfeladat 3 of the research plan.
Wearable Biomarker Signals
Mobile Content Consumption Signals
Sensor Fusion — Panic Risk Assessment
LOW RISK
On-the-Spot Intervention Protocol
Triggered automatically when panic risk exceeds threshold. Defines the sequence of AR visual + auditory neuromodulation steps.
Adjust sensor values to see intervention protocol activate.
Simulated Panic Episode Timeline
Synthetic 30-minute episode with pre-panic build-up, peak, and post-intervention decay. Shows how biomarker fusion would detect and respond in real time.
Multi-Dimensional Cancelling & Remission Flow Builder
Compose patient-specific modular treatment pipelines. Each module addresses one or more VSS symptom dimensions. Flows can be assembled, ordered, and scored against the digital twin to predict efficacy before clinical deployment.
Available Modules
Active Flow
Click modules from the palette to add them to the flow.
Symptom Dimension Coverage
Radar showing which VSS dimensions (snow, anxiety, tinnitus, aura, photopsia, cognitive, autonomic, sleep, mood, motor) are addressed by the current flow.
Build a flow first.
Cancelling vs. Remission Structures
VS Cancelling
Real-time suppression of the perceived snow field via AR phase-cancellation and Gabor wavelet filters. Provides immediate perceptual relief without altering the underlying neural state. Effective for acute exacerbations; requires continuous AR hardware.
Neuroplastic Remission
Long-term thalamocortical dysrhythmia reduction through rTMS, syntonic phototherapy, and neuro-optometric rehabilitation. Slow-acting (weeks–months) but durable without continuous hardware dependency.
Behavioural Loop Correction
Breaks the perception–anxiety–exacerbation spiral via digital twin-guided lifestyle modification (sleep hygiene, screen protocols, mindfulness) supported by the homunculus AI agent.
Panic Interrupt
On-the-spot biomarker-triggered intervention sequence. Deploys AR noise suppression + binaural audio neuromodulation within <2s of panic onset detection.
Multimodal Combination
Sequentially chains modules across all four structure types, weighted by patient-specific dimension severity. The digital twin pre-scores each combination before human trial.
Suggested Further Research Directions
EEG Neurofeedback Integration
Closed-loop thalamocortical gamma-suppression via real-time Morlet CWT + binaural beats calibrated to individual alpha peak frequency. Link EEG Analysis tab biomarkers to AR filter adjustment.
Federated Learning Cohort
Distribute model training across patient devices without centralising raw data. Enables privacy-preserving homunculus AI improvement across the 20,000+ user base.
Longitudinal Digital Phenotyping
Extend simulation exports to include passive sensing (gait, typing cadence, GPS mobility). Richer behavioural signatures improve digital twin fidelity and panic prediction accuracy.
Pharmacokinetic Twin Layer
Add a pharmacological module to the digital twin that simulates lamotrigine / acetazolamide plasma curves and their predicted effect on thalamocortical excitability within the Markov state space.
Cross-Patient Common Signal Finder
Use the cross-spectrogram from EEG Analysis across patient cohorts to identify candidate shared neural oscillatory sources — a pathway toward objective VSS biomarker discovery.
SaMD Regulatory Pathway
Define evidence package for FDA/CE Software as a Medical Device submission: clinical validation study design, IEC 62304 compliant architecture, and GDPR-aligned data governance for the Cloudflare Workers backend.
Research Sessions
| Title | Type | Status | Created | Actions |
|---|---|---|---|---|
| Loading sessions… | ||||