🏥 HealthMesh v2.0

Intelligent Clinical Decision Support Platform
A production-ready, multi-agent healthcare orchestration system powered by Microsoft Azure AI Services

Built for Microsoft Imagine Cup 2026 | Healthcare Category

---

📋 Release Notes - Version 2.0.0

Release Date: January 3, 2026

🆕 What's New in v2.0

Feature	Description
🧠 Azure OpenAI Integration	GPT-4o powered clinical reasoning with intelligent fallback
⚡ Early Deterioration Detection	AI-powered NEWS2/SOFA scoring with real-time patient monitoring
💊 Medication Safety Engine	Context-aware drug interaction detection with severity grading
🤖 Agent Orchestrator	Visual multi-agent pipeline with step-by-step execution
🔐 Microsoft Entra ID	Enterprise-grade authentication with multi-tenant isolation
📊 Azure SQL Database	Persistent cloud storage with hospital-level data isolation
🚀 Production Deployment	GitHub Actions CI/CD with Azure App Service

---

🌟 Overview

HealthMesh is an enterprise-grade, multi-tenant clinical decision support platform designed to assist healthcare professionals in making informed decisions for complex patient cases. The platform leverages Microsoft Azure AI Services to provide real-time clinical insights, evidence-based recommendations, and transparent AI reasoning.

Architecture Highlights

• Azure OpenAI: GPT-4o powered clinical intelligence with rule-based fallback
• Multi-Tenant Isolation: Complete data separation per hospital/organization
• HIPAA-Ready: Comprehensive audit logging and encryption
• FHIR R4 Compatible: Standard healthcare data interoperability

Quick Navigation

📚 Complete Documentation | 🔒 Security Policy | 🚀 Deployment Guide | 📁 Project Structure

Security Features:

• Security Scorecard (99/100) ⭐
• RBAC Authorization
• Field Encryption
• Vulnerability Report

---

✨ Complete Feature List

🤖 Multi-Agent Clinical Intelligence Pipeline (6 Agents)

Agent	Role	Key Outputs
🚨 Triage Agent	NEWS2/SOFA scoring, risk classification	Urgency score, risk category, red flags
🩺 Diagnostic Agent	Differential diagnosis generation	Ranked diagnoses with probability scores
📋 Guideline Agent	Medical guideline mapping (NCCN, WHO, ICMR, ADA)	Evidence-based recommendations
💊 Medication Safety Agent	Drug interactions, dosing, allergies	Severity-graded alerts with clinical context
📚 Evidence Agent	RAG-powered clinical research retrieval	PubMed/guideline citations
🧠 Synthesis Orchestrator	Multi-agent output integration	Unified clinical summary

💊 Medication Safety Engine (NEW in v2.0)

Feature	Description
Drug-Drug Interactions	15+ interaction rules with clinical mechanisms
Context-Aware Severity	Age, renal function modify alert severity
Allergy Cross-Reactivity	Detects related drug class allergies
Renal/Hepatic Dosing	Flags medications needing adjustment
Explainability Panel	Data sources, reasoning chain, limitations
Alert Fatigue Prevention	Suppressible low-priority alerts

Detected Interactions Include:

• Aspirin + ACE Inhibitors → Reduced cardioprotection
• Metformin + Sulfonylureas → Hypoglycemia risk
• Statins + Amlodipine → Myopathy risk
• NSAIDs + Anticoagulants → Bleeding risk
• And many more...

⚡ Early Deterioration Detection (NEW in v2.0)

Feature	Description
NEWS2 Scoring	National Early Warning Score 2 calculation
SOFA-lite Assessment	Organ dysfunction evaluation
AI Enhancement	Azure OpenAI-powered trend analysis
Real-Time Signals	Vital sign deterioration tracking
Clinical Recommendations	Actionable intervention suggestions
Risk Trajectory	Predictive deterioration modeling

🎭 Agent Orchestrator (Visual Pipeline)

• Step-by-Step Execution: Watch each agent process in real-time
• Dependency Resolution: Automatic agent sequencing
• Parallel Execution: Independent agents run simultaneously
• Output Visualization: Expandable results per agent
• Error Handling: Graceful fallbacks on agent failure

📱 Patient QR Identity System

• FHIR R4 Compliant Patient Master ID
• Unique QR Codes for instant patient identification
• Mobile Scanning with role-based access control
• Longitudinal Dashboard on scan

🔐 Enterprise Security & Multi-Tenancy

Feature	Implementation
Authentication	Microsoft Entra ID (Azure AD) only
Data Isolation	Hospital-level tenant separation
Access Control	Role-Based (Doctor, Nurse, Admin)
Audit Logging	Full action trail with timestamps
Zero Passwords	No local credential storage

📊 Clinical Scoring Algorithms

• NEWS2 (National Early Warning Score 2)
• SOFA-lite (Sequential Organ Failure Assessment)
• Risk Classification: Low / Moderate / High / Critical

🔍 Explainable AI

• Step-by-step reasoning chains
• Confidence scores per agent
• Data source citations
• Known limitations disclosure
• Clinical disclaimers

---

🏗️ System Architecture

┌─────────────────────────────────────────────────────────────────────────────┐
│                         HEALTHMESH v2.0 PLATFORM                            │
├─────────────────────────────────────────────────────────────────────────────┤
│                                                                             │
│  ┌─────────────────────┐    ┌─────────────────────┐    ┌─────────────────┐ │
│  │     FRONTEND        │    │      BACKEND        │    │   AZURE AI      │ │
│  │                     │    │                     │    │   SERVICES      │ │
│  │  • React 18 + TS    │    │  • Express.js       │    │                 │ │
│  │  • Tailwind CSS     │◄──►│  • TypeScript       │◄──►│ 🔵 Azure OpenAI │ │
│  │  • Radix UI         │    │  • Node.js 22       │    │    (GPT-4o)     │ │
│  │  • TanStack Query   │    │  • RESTful APIs     │    │ 📋 Rule Fallback│ │
│  │  • Wouter Router    │    │  • Multi-Agent      │    │                 │ │
│  └─────────────────────┘    └─────────────────────┘    └─────────────────┘ │
│                                       │                                     │
│  ┌────────────────────────────────────┴────────────────────────────────┐   │
│  │                        AZURE SERVICES LAYER                          │   │
│  │                                                                      │   │
│  │  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐  ┌──────────┐ │   │
│  │  │ Azure SQL    │  │ Microsoft    │  │ Azure        │  │ Azure    │ │   │
│  │  │ Database     │  │ Entra ID     │  │ App Service  │  │ Monitor  │ │   │
│  │  │              │  │              │  │              │  │          │ │   │
│  │  │ Multi-tenant │  │ Enterprise   │  │ Production   │  │ Logging  │ │   │
│  │  │ Persistence  │  │ Auth/SSO     │  │ Hosting      │  │ & Audit  │ │   │
│  │  └──────────────┘  └──────────────┘  └──────────────┘  └──────────┘ │   │
│  │                                                                      │   │
│  │  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐  ┌──────────┐ │   │
│  │  │ Azure OpenAI │  │ Azure Health │  │ Azure        │  │ Azure    │ │   │
│  │  │ (GPT-4o)     │  │ Data Services│  │ Cognitive    │  │ Document │ │   │
│  │  │              │  │ (FHIR R4)    │  │ Search       │  │ Intel    │ │   │
│  │  │ AI Engine    │  │ HL7 FHIR API │  │ RAG Index    │  │ Lab OCR  │ │   │
│  │  └──────────────┘  └──────────────┘  └──────────────┘  └──────────┘ │   │
│  └─────────────────────────────────────────────────────────────────────┘   │
│                                                                             │
└─────────────────────────────────────────────────────────────────────────────┘

---

🛠️ Technology Stack

Frontend

Technology	Version	Purpose
React	18.x	UI Framework
TypeScript	5.x	Type Safety
Tailwind CSS	3.x	Styling
Radix UI	Latest	Accessible Components
TanStack Query	5.x	Data Fetching & Caching
Wouter	3.x	Client-side Routing
Recharts	2.x	Data Visualization
Framer Motion	11.x	Animations

Backend

Technology	Version	Purpose
Express.js	4.x	Web Framework
TypeScript	5.x	Type Safety
Node.js	22.x LTS	Runtime
mssql	11.x	Azure SQL Client
Zod	3.x	Schema Validation
MSAL Node	2.x	Azure AD Authentication

Azure Services

Service	Purpose
Azure OpenAI (GPT-4o)	Clinical reasoning & analysis
Azure SQL Database	Multi-tenant data persistence
Microsoft Entra ID	Enterprise authentication
Azure App Service	Production web hosting
Azure Monitor	Logging & telemetry
Azure Health Data Services	FHIR R4 (optional)
Azure Document Intelligence	Lab report OCR
Azure Cognitive Search	Medical guidelines RAG
Azure Key Vault	Secrets management
Azure Application Insights	Performance monitoring

---

📁 Project Structure

healthmesh/
├── client/                          # Frontend React Application
│   ├── src/
│   │   ├── pages/                   # Route Pages
│   │   │   ├── dashboard.tsx        # Main dashboard
│   │   │   ├── patients.tsx         # Patient management
│   │   │   ├── cases.tsx            # Clinical cases
│   │   │   ├── orchestrator.tsx     # Agent Orchestrator (NEW)
│   │   │   ├── early-deterioration.tsx  # NEWS2/SOFA (NEW)
│   │   │   ├── medication-safety.tsx    # Drug safety (NEW)
│   │   │   ├── labs.tsx             # Lab reports
│   │   │   ├── qr-scan.tsx          # QR scanner
│   │   │   ├── audit.tsx            # Audit logs
│   │   │   └── settings.tsx         # User settings
│   │   ├── components/              # Reusable UI Components
│   │   ├── auth/                    # MSAL Authentication
│   │   └── lib/                     # Utilities
│   └── index.html
│
├── server/                          # Backend Express Server
│   ├── agents/                      # AI Agent Engines (NEW)
│   │   ├── ai-medication-safety-agent.ts   # AI medication safety
│   │   ├── medication-safety-engine.ts     # Rule-based drug interactions
│   │   ├── early-deterioration-agent.ts    # NEWS2/SOFA scoring
│   │   └── ai-enhanced-deterioration-agent.ts  # AI deterioration
│   ├── azure/                       # Azure Service Clients
│   │   ├── openai-client.ts         # Azure OpenAI
│   │   ├── monitoring.ts            # Azure Monitor
│   │   └── config.ts                # Configuration
│   ├── auth/                        # Authentication
│   │   └── entraAuth.ts             # Microsoft Entra ID
│   ├── api/                         # API Route Modules
│   │   ├── medication-safety-routes.ts  # Medication safety API
│   │   └── deterioration-routes.ts      # Deterioration API
│   ├── db/                          # Database Layer
│   │   ├── azure-sql.ts             # Azure SQL connection
│   │   └── migrations/              # SQL migrations
│   ├── clinical-agents.ts           # Core clinical pipeline
│   ├── api-routes.ts                # Main API routes
│   └── index.ts                     # Server entry point
│
├── shared/                          # Shared TypeScript Types
│   └── schema.ts                    # Zod Schemas
│
├── docs/                            # Documentation
│   ├── EARLY_DETERIORATION_AGENT.md
│   ├── AGENT_PROMPTS.md
│   └── ...
│
├── .github/workflows/               # CI/CD Pipelines
│   └── deploy-production.yml
│
└── infra/                           # Azure Infrastructure (Bicep)
    └── main.bicep

---

🚀 Quick Start

Prerequisites

• Node.js 22 LTS (Download)
• Azure Subscription (Create free account)
• Azure CLI (Install)

1. Clone the Repository

git clone https://github.com/balaraj74/HealthMesh.git
cd HealthMesh

2. Install Dependencies

npm install

3. Configure Environment

cp .env.azure.example .env

Edit .env with your Azure credentials:

# ===== Azure OpenAI =====
AZURE_OPENAI_ENDPOINT=https://your-openai.openai.azure.com/
AZURE_OPENAI_API_KEY=your-api-key
AZURE_OPENAI_DEPLOYMENT=gpt-4o

# ===== Azure Authentication =====
AZURE_AD_TENANT_ID=your-tenant-id
AZURE_AD_CLIENT_ID=your-client-id
VITE_AZURE_AD_TENANT_ID=your-tenant-id
VITE_AZURE_AD_CLIENT_ID=your-client-id

# ===== Azure SQL Database =====
AZURE_SQL_CONNECTION_STRING=Server=your-server.database.windows.net;Database=healthmesh;User Id=admin;Password=password;Encrypt=true

# ===== Optional Services =====
AZURE_FHIR_ENDPOINT=https://your-fhir.fhir.azurehealthcareapis.com
AZURE_DOCUMENT_INTELLIGENCE_ENDPOINT=https://your-docint.cognitiveservices.azure.com/
AZURE_DOCUMENT_INTELLIGENCE_KEY=your-key
AZURE_SEARCH_ENDPOINT=https://your-search.search.windows.net
AZURE_SEARCH_ADMIN_KEY=your-key

4. Start Development Server

npm run dev

Application available at http://localhost:5000

5. Sign In

1. Open http://localhost:5000
2. Click "Sign In with Microsoft"
3. Authenticate with your Azure AD credentials

---

📡 API Reference

Authentication

All endpoints require Microsoft Entra ID Bearer token:

Authorization: Bearer <entra-access-token>

Core Endpoints

Method	Endpoint	Description
GET	/api/me	Get authenticated user profile
GET	/api/patients	List patients (hospital-scoped)
POST	/api/patients	Create new patient
GET	/api/cases	List clinical cases
POST	/api/cases/:id/clinical-analyze	Run 6-agent analysis

Medication Safety API (NEW)

Method	Endpoint	Description
POST	/api/medication-safety/analyze-real	AI-powered safety analysis
POST	/api/medication-safety/check-new	Check new medication
POST	/api/medication-safety/alert/:id/acknowledge	Acknowledge alert
GET	/api/medication-safety/patient/:id/history	Alert history

Early Deterioration API (NEW)

Method	Endpoint	Description
POST	/api/deterioration/analyze	Basic NEWS2/SOFA
POST	/api/deterioration/analyze-real	AI-enhanced analysis
GET	/api/deterioration/patient/:id/alerts	Patient alerts

Lab Reports API

Method	Endpoint	Description
GET	/api/lab-reports/patient/:id	Get patient lab reports
POST	/api/lab-reports/upload	Upload lab report (OCR)

QR Identity API

Method	Endpoint	Description
GET	/api/qr/patient/:id	Get patient QR code
POST	/api/qr/scan	Scan QR and retrieve patient

---

🔒 Security & Compliance

Authentication Flow

User → Microsoft Entra ID → JWT Token → Backend Validation → Hospital Context

Compliance Features

Standard	Implementation
HIPAA Ready	Audit logging, encryption, access controls
FHIR R4	Azure Health Data Services integration
SOC 2 Ready	Azure security controls
GDPR Compliant	Data isolation, consent management

Security Controls

• ✅ HTTPS-only with TLS 1.2+
• ✅ Role-Based Access Control (RBAC)
• ✅ Hospital-level data isolation
• ✅ Full audit trail logging
• ✅ No local password storage
• ✅ Managed identities for Azure services
• ✅ Azure Key Vault for secrets

---

🚢 Production Deployment

GitHub Actions CI/CD

The repository includes automated deployment:

# .github/workflows/deploy-production.yml
# Triggers on push to main branch
# Deploys to Azure App Service

Manual Deployment

# Build application
npm run build

# Deploy to Azure App Service
az webapp deployment source config-zip \
  --resource-group healthmesh-rg \
  --name healthmesh-app \
  --src deploy.zip

Azure Infrastructure (Bicep)

# Deploy infrastructure
az deployment group create \
  --resource-group healthmesh-rg \
  --template-file infra/main.bicep \
  --parameters environment=production

---

📊 Monitoring & Observability

Azure Monitor Integration

• Application Insights: Performance metrics, distributed tracing
• Log Analytics: Centralized logging
• Azure Alerts: Proactive issue detection

Tracked Metrics

Metric	Description
Agent Execution Time	Per-agent performance
AI Token Usage	Azure OpenAI consumption
Confidence Scores	Analysis reliability
Alert Generation	Safety alerts created
API Response Times	Endpoint latency
Error Rates	System health indicators

---

🧪 Available Scripts

Command	Description
npm run dev	Start development server (port 5000)
npm run build	Build for production
npm start	Run production server
npm run check	TypeScript type checking

---

⚠️ Clinical Disclaimer

This system is designed as CLINICAL DECISION SUPPORT only.

• All recommendations must be reviewed by a licensed clinician
• The AI does NOT make diagnoses or treatment decisions
• The final clinical decision rests with the treating physician
• Evidence and guidelines may change; always verify current standards
• Patient-specific factors may override general recommendations

---

🏆 Microsoft Imagine Cup 2026

HealthMesh is built for Microsoft Imagine Cup 2026 in the Healthcare category.

Judging Criteria Alignment

Criteria	Implementation
Technical Excellence	Multi-agent AI architecture with Azure services
Innovation	6-agent explainable clinical AI pipeline
Impact	Addresses medication safety and deterioration detection
Azure Integration	Deep integration with 10+ Azure services
Production Ready	Complete with CI/CD, monitoring, and multi-tenancy

Key Differentiators

• Multi-Agent Architecture: 6 specialized clinical AI agents
• Explainable AI: Full transparency in reasoning
• Enterprise Ready: Multi-tenant with hospital isolation
• Real Clinical Value: NEWS2, SOFA, drug interaction detection

---

👨‍💻 Author & Contact

Balaraj R
Developer & Owner

• 📧 Email: balarajr483@gmail.com
• 💼 LinkedIn: linkedin.com/in/balaraj-r-209a67330
• 🐙 GitHub: github.com/balaraj74

---

📜 License

This project is proprietary software developed for Microsoft Imagine Cup 2026. All rights reserved © 2026 Balaraj R.

---

🙏 Acknowledgments

• Microsoft Azure team for comprehensive cloud services
• Microsoft Imagine Cup for the opportunity
• Healthcare professionals who provided clinical insights
• Open-source community for foundational technologies

---

Made with ❤️ for better healthcare
Powered by Microsoft Azure AI Services