ParkinGuide is an original digital health product designed for longitudinal functional intelligence, contextual monitoring, and explainable functional risk analysis in Parkinson’s disease.
Unlike traditional healthcare systems centered on isolated clinical snapshots, ParkinGuide focuses on what happens between consultations, transforming real-life motor, emotional, social and cognitive signals into structured longitudinal insights.
It is not a diagnostic system.
It does not recommend treatments, adjust medication, or replace healthcare professionals.
Its purpose is to support better-informed neurological follow-up through structured longitudinal observation and explainable functional interpretation.
Parkinson’s progression is highly heterogeneous.
Two patients with the same diagnosis may evolve in completely different ways.
Population averages are often insufficient to understand functional deterioration.
ParkinGuide follows a precision longitudinal health approach:
Each patient is interpreted against:
their own baseline
their own trajectory
their own functional context
Not against population averages.
This enables:
personalized functional interpretation over time
Main product vision and positioning.
Longitudinal intelligence over episodic care.
Population-level contextual monitoring.
Includes:
- functional trends
- contextual signals
- cognitive activity
- wearable-ready variables
- recent clinical activity
Personalized 24-week functional evolution.
Shows baseline-relative progression and contextual changes over time.
Higher-risk progression profile showing accumulated contextual deterioration.
Integrated mini-games for repeated functional observation.
Used to monitor cognitive speed, motor precision and executive activation.
Example of a cognitive-motor task used to monitor executive function and language activation longitudinally.
Structured longitudinal PDF reports designed to improve neurologist-patient consultations.
ParkinGuide is built to:
- record real-world contextual variables
- detect functional deviations over time
- identify deterioration patterns
- reduce information loss between consultations
- improve continuity of care
- generate structured clinical summaries
- support explainable AI in personalized neurological care
ParkinGuide is built around three complementary intelligence layers.
The system compares:
Current patient state
vs
their own historical baseline
Never against other patients.
Example:
5,000 steps/day
may represent:
- normal stability for one patient
- significant decline for another
This enables:
personalized functional interpretation
ParkinGuide supports:
- patient self-report
- caregiver observation
- combined reporting
This allows comparison between:
perceived function
vs
observed function
These discrepancies may become clinically meaningful contextual signals.
Synthetic longitudinal dataset:
100 patients
24 weeks each
2400 weekly records
6000 cognitive-motor results
Used to model:
- nonlinear deterioration patterns
- variable interactions
- contextual instability
- habit-response dynamics
This layer enriches patient-level baseline modelling.
ParkinGuide monitors 21 weekly functional variables across four complementary domains.
Tracks physical and motor performance:
- tremor
- rigidity
- bradykinesia
- freezing
- falls
- balance confidence
- steps
- training hours
Tracks emotional regulation and functional self-perception:
- stress
- mood
- motivation
- autonomy
- mental fatigue
Tracks environmental and contextual variables:
- sleep hours
- sleep quality
- night awakenings
- social interactions
- social isolation
- hydration
- constipation
Measured through integrated mini-games:
- verbal fluency
- reaction time
- memory
- coordination
- executive activation
These allow repeated observation of subtle functional changes over time.
Integrated assessments:
- Visual Memory
- Reaction Time
- Manual Coordination
- Verbal Fluency
- Number Sequence
Used to monitor:
- executive function
- cognitive speed
- motor precision
- memory evolution
- repeated functional patterns
ParkinGuide integrates two complementary AI systems.
Rule-based layer.
Examples:
low sleep + high fatigue
high freezing + low balance confidence
high stress + low activity
Produces:
ContextInsight
Properties:
- transparent
- explainable
- auditable
- non-diagnostic
Current model:
Random Forest Regressor
Predicts:
functional_risk_score (0–100)
Why Random Forest:
- nonlinear modeling
- robust with mixed variables
- stable for medium datasets
- interpretable feature importance
- handles interaction effects well
MAE = 3.59
RMSE = 4.84
R² = 0.946
Top predictors:
- Balance confidence
- Sleep hours
- Freezing episodes
- Bradykinesia
- Sleep quality
- Daily steps
- Social isolation
- Mental fatigue
This shows the model is already learning:
which variables tend to drive functional deterioration
Patient
↓
Weekly records
↓
Contextual AI
↓
Functional risk prediction
↓
Longitudinal visualization
↓
Clinical PDF report
↓
Neurology consultation
Outputs:
- contextual summary
- symptom evolution
- functional risk score
- cognitive history
- consultation-ready reports
Patient
↓
WeeklyRecord
↓
Longitudinal Database
↓
Context Engine
↓
Random Forest Predictor
↓
ContextInsight
↓
Longitudinal Visualization
↓
Clinical Report
Built with:
- pseudonymized IDs
- no real health identifiers
- explainable AI layers
- transparent logic
- responsible non-diagnostic positioning
Core principles:
- safety
- explainability
- responsibility
- clinical usefulness
- Python
- Django
- SQLite
- PostgreSQL-ready architecture
- HTML
- CSS
- JavaScript
- Bootstrap
- Chart.js
- Pandas
- NumPy
- Scikit-learn
- ReportLab
- Longitudinal patient profiles
- Weekly contextual records
- Cognitive-motor tracking
- Functional dashboard
- Clinical PDF reports
- Synthetic dataset generation
- Functional risk engineering
- Random Forest model
- Real-time prediction
- Context insight generation
- Longitudinal risk visualization
- Alert engine
- Wearable synchronization
- Per-patient explainability
- AI-generated consultation summaries
- Habit-response correlation engine
ParkinGuide does not aim to diagnose disease.
Its value is more clinically meaningful:
- understand how a patient is evolving
- identify what may be influencing deterioration
- detect which variables matter most
- reduce information loss between consultations
- improve the quality of clinical conversations
That makes it:
a functional intelligence layer for longitudinal neurological care
Bea Lamiquiz
Backend Developer · Applied AI Product Developer · Digital Health Systems
GitHub: https://github.com/beatriangu
LinkedIn: https://www.linkedin.com/in/beatrizlamiquiz
Developed within:
Laborlan — Artificial Intelligence and Technological Projects
Focus areas:
- Applied AI
- Product thinking
- Explainable systems
- Precision health
- Longitudinal analytics
ParkinGuide is an original proprietary digital health framework currently under active development.
Its conceptual architecture, longitudinal modelling methodology, contextual intelligence system and functional risk logic are original work by the author.
© 2026 Beatriz Lamiquiz. All rights reserved.