test/VAULTMESH-ETERNAL-PATTERN/funding-roadmap/diagrams/genai-health-federated-governance.mmd

%% VaultMesh GenAI Health — Federated Learning + Governance Flow
%% Proposal: €3M HORIZON-HLTH-2025-CARE-01
%% Partners: VaultMesh (IE), DFKI (DE), UMC Utrecht (NL), Philips Healthcare (NL)

graph TB
    subgraph DataSources["🏥 Federated Data Sources (Privacy-Preserving)"]
        HOSPITAL1[Hospital A<br/>Electronic Health Records<br/>Germany — 10k patients]
        HOSPITAL2[Hospital B<br/>Wearable + Lab Data<br/>Netherlands — 8k patients]
        HOSPITAL3[Hospital C<br/>Imaging + Genomics<br/>France — 5k patients]
        CLINIC[Primary Care Clinics<br/>Long-Term Health Monitoring<br/>Multi-Country — 20k patients]
    end

    subgraph Privacy["🔒 Privacy & Anonymization Layer"]
        ANON[Data Anonymization<br/>Differential Privacy (ε=0.1)<br/>K-Anonymity (k≥10)]
        ENCRYPT[Homomorphic Encryption<br/>Computation on Encrypted Data<br/>SEAL/Palisade Library]
        CONSENT[Consent Management<br/>GDPR Art. 7 & 9 Compliance<br/>Dynamic Consent Updates]

        WP2_PRIVACY[WP2: Privacy Infrastructure<br/>Lead: DFKI<br/>M1-12]
    end

    subgraph FederatedML["🤖 Federated Machine Learning"]
        LOCAL_TRAIN1[Local Training — Site A<br/>No Raw Data Leaves Hospital<br/>Model Gradients Only]
        LOCAL_TRAIN2[Local Training — Site B<br/>Secure Aggregation Protocol<br/>Encrypted Weight Updates]
        LOCAL_TRAIN3[Local Training — Site C<br/>Edge Computing<br/>GPU Clusters On-Premise]

        AGGREGATOR[Secure Aggregator<br/>Federated Averaging (FedAvg)<br/>Byzantine-Robust]
        GLOBAL_MODEL[Global Model<br/>Disease Prediction (Diabetes, CVD)<br/>Treatment Recommendation]

        WP3_ML[WP3: Federated Learning<br/>Lead: UMC Utrecht<br/>M4-18]
    end

    subgraph Governance["⚖️ AI Act Governance & Human-in-the-Loop"]
        PSI[Ψ-Field Oversight<br/>Collective Sensing of Model Drift<br/>Bias Detection & Alerts]
        ETHICS_BOARD[Ethics Review Board<br/>Continuous Monitoring<br/>IRB Integration]
        HUMAN_REVIEW[Human-in-the-Loop<br/>Clinician Review of High-Risk Decisions<br/>AI Act Art. 14 Compliance]
        EXPLAINABLE[Explainability Layer<br/>SHAP/LIME for Model Interpretation<br/>Clinician-Facing Dashboard]

        WP4_GOV[WP4: Governance & Compliance<br/>Lead: VaultMesh<br/>M6-24]
    end

    subgraph VaultMesh["🏛️ VaultMesh Proof Infrastructure"]
        RECEIPTS[Receipt Engine<br/>Every Training Round = Cryptographic Receipt<br/>Provenance Trail]
        LAWCHAIN[LAWCHAIN<br/>Model Version History<br/>Immutable Audit Log]
        FEDERATION[Federation Router<br/>Hospital-to-Hospital Secure Channel<br/>mTLS + Capability Exchange]
        TREASURY[Treasury Subsystem<br/>Computation Cost Tracking<br/>Fair Resource Allocation]

        WP1[WP1: Core Integration<br/>Lead: VaultMesh<br/>M1-6]
    end

    subgraph Deployment["🚀 Clinical Deployment Pipeline"]
        VALIDATION[Clinical Validation<br/>Retrospective Study (n=5000)<br/>Prospective Trial (n=500)]
        REGULATORY[Regulatory Pathway<br/>CE Mark (MDR)<br/>FDA Pre-Cert if US expansion]
        INTEGRATION[EHR Integration<br/>HL7 FHIR API<br/>Philips IntelliSpace]

        PILOT[🧪 UMC Utrecht Pilot<br/>Diabetes Risk Prediction<br/>12-Month Trial — M12-24]

        WP5_PILOT[WP5: Pilots & Validation<br/>Lead: UMC Utrecht<br/>M12-24]
    end

    subgraph Policy["🇪🇺 EU Policy & Standards Compliance"]
        AI_ACT[AI Act (Reg 2024/1689)<br/>Art. 10: High-Risk System Data Governance<br/>Art. 14: Human Oversight]
        GDPR[GDPR (Art. 9)<br/>Special Category Health Data<br/>Explicit Consent + Safeguards]
        MDR[Medical Device Reg (MDR)<br/>Class IIa/IIb Risk Classification<br/>Clinical Evidence Requirements]
        EHDS[European Health Data Space<br/>Data Altruism + Secondary Use<br/>Cross-Border Access]
    end

    subgraph Outcomes["📊 Expected Outcomes & KPIs"]
        KPI1[20% Diagnostic Accuracy Improvement<br/>vs. Current Clinical Baselines]
        KPI2[30% Reduction in Data Collection Time<br/>Federated vs. Centralized]
        KPI3[5+ Hospital Network<br/>Cross-Border Federation Active]
        KPI4[0 Privacy Breaches<br/>Differential Privacy Guarantees]
        KPI5[10+ Publications<br/>Top-Tier Medical AI Venues]
    end

    %% Data → Privacy Flow
    HOSPITAL1 --> ANON
    HOSPITAL2 --> ANON
    HOSPITAL3 --> ANON
    CLINIC --> ANON
    ANON --> ENCRYPT
    ENCRYPT --> CONSENT
    CONSENT -.->|"Audit Trail"| RECEIPTS

    %% Privacy → Federated ML
    CONSENT --> LOCAL_TRAIN1
    CONSENT --> LOCAL_TRAIN2
    CONSENT --> LOCAL_TRAIN3
    LOCAL_TRAIN1 -->|"Encrypted Gradients"| AGGREGATOR
    LOCAL_TRAIN2 -->|"Encrypted Gradients"| AGGREGATOR
    LOCAL_TRAIN3 -->|"Encrypted Gradients"| AGGREGATOR
    AGGREGATOR --> GLOBAL_MODEL

    %% Federated ML → Governance
    GLOBAL_MODEL -->|"Model Artifact"| PSI
    GLOBAL_MODEL -->|"Prediction Requests"| HUMAN_REVIEW
    PSI -->|"Drift Alerts"| ETHICS_BOARD
    ETHICS_BOARD -->|"Review Decision"| HUMAN_REVIEW
    HUMAN_REVIEW --> EXPLAINABLE
    EXPLAINABLE -.->|"Clinician Feedback"| GLOBAL_MODEL

    %% VaultMesh Integration
    RECEIPTS -->|"Training Receipt"| AGGREGATOR
    RECEIPTS -->|"Governance Decision Receipt"| HUMAN_REVIEW
    LAWCHAIN -->|"Model Version Chain"| GLOBAL_MODEL
    LAWCHAIN -->|"Audit Queries"| ETHICS_BOARD
    FEDERATION -->|"Hospital Network"| AGGREGATOR
    TREASURY -->|"Computation Metering"| LOCAL_TRAIN1
    TREASURY -->|"Computation Metering"| LOCAL_TRAIN2
    TREASURY -->|"Computation Metering"| LOCAL_TRAIN3

    %% Deployment Pipeline
    GLOBAL_MODEL --> VALIDATION
    VALIDATION --> REGULATORY
    REGULATORY --> INTEGRATION
    INTEGRATION --> PILOT
    PILOT -.->|"Clinical Feedback"| GLOBAL_MODEL

    %% Policy Compliance
    CONSENT -->|"GDPR Art. 9"| GDPR
    PSI -->|"AI Act Art. 14"| AI_ACT
    HUMAN_REVIEW -->|"AI Act Art. 14"| AI_ACT
    VALIDATION -->|"MDR Evidence"| MDR
    FEDERATION -->|"EHDS Cross-Border"| EHDS
    ANON -->|"GDPR Art. 25 (Privacy by Design)"| GDPR

    %% Work Package Dependencies
    WP1 --> RECEIPTS
    WP1 --> LAWCHAIN
    WP1 --> FEDERATION
    WP2_PRIVACY --> ANON
    WP2_PRIVACY --> ENCRYPT
    WP2_PRIVACY --> CONSENT
    WP3_ML --> AGGREGATOR
    WP3_ML --> GLOBAL_MODEL
    WP4_GOV --> PSI
    WP4_GOV --> ETHICS_BOARD
    WP4_GOV --> HUMAN_REVIEW
    WP5_PILOT --> PILOT
    WP5_PILOT --> VALIDATION

    %% Ψ-Field Collective Intelligence
    PSI -.->|"Cross-Site Anomaly Patterns"| AGGREGATOR
    LOCAL_TRAIN1 -.->|"Site A Metrics"| PSI
    LOCAL_TRAIN2 -.->|"Site B Metrics"| PSI
    LOCAL_TRAIN3 -.->|"Site C Metrics"| PSI

    %% Outcomes Tracking
    GLOBAL_MODEL -->|"Accuracy Metrics"| KPI1
    AGGREGATOR -->|"Training Efficiency"| KPI2
    FEDERATION -->|"Network Growth"| KPI3
    ENCRYPT -->|"Privacy Guarantees"| KPI4
    VALIDATION -->|"Research Output"| KPI5

    classDef data fill:#e3f2fd,stroke:#01579b,stroke-width:2px
    classDef privacy fill:#f3e5f5,stroke:#6a1b9a,stroke-width:2px
    classDef ml fill:#c8e6c9,stroke:#2e7d32,stroke-width:2px
    classDef governance fill:#fff9c4,stroke:#f57f17,stroke-width:3px
    classDef vaultmesh fill:#ffccbc,stroke:#bf360c,stroke-width:2px
    classDef deployment fill:#b2dfdb,stroke:#004d40,stroke-width:2px
    classDef policy fill:#e1bee7,stroke:#4a148c,stroke-width:2px
    classDef outcomes fill:#ffe0b2,stroke:#e65100,stroke-width:2px

    class HOSPITAL1,HOSPITAL2,HOSPITAL3,CLINIC data
    class ANON,ENCRYPT,CONSENT privacy
    class LOCAL_TRAIN1,LOCAL_TRAIN2,LOCAL_TRAIN3,AGGREGATOR,GLOBAL_MODEL ml
    class PSI,ETHICS_BOARD,HUMAN_REVIEW,EXPLAINABLE governance
    class RECEIPTS,LAWCHAIN,FEDERATION,TREASURY vaultmesh
    class VALIDATION,REGULATORY,INTEGRATION,PILOT deployment
    class AI_ACT,GDPR,MDR,EHDS policy
    class KPI1,KPI2,KPI3,KPI4,KPI5 outcomes