HYGIEIA — Multi-Disease Diagnosis Framework

5Diagnostic models

96%+Average accuracy

253KTraining samples

4Security layers

01The gist

What it does

Hygieia bundles five independently trained diagnostic models behind one interface, so a clinician or a curious patient can move between a heart-risk questionnaire and a dermatology image scan without changing tools.

Every prediction is written to an append-only chain with a SHA-256 hash, which means a result can be re-verified later and any tampering shows up as a broken chain. On top of that sits Dr. Hygieia, a context-aware assistant that can read your own analysis history when it answers.

5 AI models
Specialised diagnostic and predictive models, each trained and tuned separately.
96%+ accuracy
Clinical-grade predictions validated with ROC-AUC alongside raw accuracy.
Blockchain verified
Every analysis is hashed and chained, so records can be re-verified later.
AI health assistant
Dr. Hygieia answers with your analysis history in context, not in a vacuum.
Dark / light mode
A calm, accessible interface built for long clinical sessions.
Responsive by default
The same flows work on a ward tablet and on a phone in a waiting room.

02Model portfolio

Five models, one interface

Each model was trained, tuned and scored on its own dataset. Accuracy alone hides class imbalance, so ROC-AUC is reported next to it.

Heart Risk
18 clinical parameters
99.4%ROC-AUC 99.9%
Stacking ensemble · 303 samples
Diabetes Risk
Symptom-based analysis
98.1%ROC-AUC 99.6%
Random Forest + XGBoost · 520 samples
Skin Diagnosis
AI image analysis
96.8%ROC-AUC 99.3%
CNN + Derm Foundation · ISIC + HAM10000
Breast Cancer
10 risk factors
81.3%ROC-AUC 86.2%
Voting ensemble · 251,661 samples
Breast Tissue
30 FNA measurements
97.2%ROC-AUC 99.7%
Stacking ensemble · 569 samples

Model	Accuracy	ROC-AUC	Samples	Architecture
Heart Risk	99.4%	99.9%	303	Stacking Ensemble
Diabetes Risk	98.1%	99.6%	520	Random Forest + XGBoost
Skin Diagnosis	96.8%	99.3%	ISIC + HAM10000	CNN + Derm Foundation
Breast Cancer	81.3% (flagged: below 90%)	86.2% (flagged: below 90%)	251,661	Voting Ensemble
Breast Tissue	97.2%	99.7%	569	Stacking Ensemble

03Under the hood

How a prediction is made

Heart Risk

18 clinical parameters · Stacking Ensemble · 303 samples

Clinical input18 clinical parameters: age, cholesterol, blood pressure and more
Feature engineeringNormalisation, encoding and feature selection
Base learnersRandom Forest + XGBoost + Gradient Boosting
Meta-learnerLogistic regression stacking ensemble
Binary outputHeart disease risk prediction with a confidence score

Input parameters

Age
Sex
Chest pain type
Resting BP
Cholesterol
Fasting blood sugar
Resting ECG
Max heart rate
Exercise angina
ST depression
ST slope
Major vessels
Thalassemia

04Interface

Every screen, in context

Captured from the running application with demo data. Use the screens switch to see the light or dark theme.

Obs. 01
Guided input
Context-aware fields with real-time validation and helper text for every medical parameter. Skin lesions are uploaded as images; the other models take questionnaires and lab values.
- Toggles and units tuned per parameter
- Image upload with progress states
- A risk assessment, not a diagnosis, stated on every form
Heart risk analysis form
Obs. 02
Result, explained
Every result shows the risk level, a confidence score, the exact inputs used, the model and method, and its blockchain transaction hash. Dr. Hygieia adds a plain-language reading of what the numbers mean.
- Risk level and confidence up front
- Inputs and model details for auditability
- AI summary generated for the result
Heart risk prediction result
Obs. 03
Dr. Hygieia
A context-aware assistant that answers with your own analysis history in view. Multiple concurrent conversations, streaming responses and direct links into specific results.
- Streaming answers
- Session management
- Deep links to analyses
Dr. Hygieia chat
Obs. 04
Blockchain verification
Every record is hashed with SHA-256 and appended to a chain. Verification walks the chain, flags any tampering and gives administrators an oversight dashboard.
- Immutable audit trail
- Tamper detection and chain validation
- Admin oversight and user management
Blockchain verification dashboard