🏥 Clinical NLP Analyzer

🚀 AI-Powered Healthcare Text Intelligence Platform

👤 Author: Sabin Shrestha
🛠️ Built with: Azure AI • FastAPI • Streamlit • Docker
📅 Date: March 26, 2026 🏷️ Version: v1.0.0

---

🔥 Overview

Healthcare systems generate massive amounts of unstructured clinical data.

👉 This application transforms that into:

• Structured clinical data
• Standardized medical coding
• Actionable insights

---

🖥️ Application Preview

🧠 Dashboard Overview

---

📊 Extraction Results

---

✨ Key Features

• 🩺 Extract Diagnoses, Medications, Symptoms
• 🔗 Map to ICD-10, SNOMED CT, UMLS, DrugBank
• 🧠 AI-powered clinical understanding
• 📊 Clean dashboard UI
• 📥 Downloadable reports
• 🐳 Fully Dockerized (production-ready)

---

🧠 Architecture

[ Streamlit UI ]  →  [ FastAPI Backend ]  →  [ Azure Healthcare NLP ]
         (8501)              (8000)                  (5000)

---

🛠️ Tech Stack

Layer	Technology
Frontend	Streamlit
Backend	FastAPI
AI/NLP	Azure Text Analytics for Health
Deployment	Docker Compose

---

⚙️ Prerequisites

• Docker Desktop (running)
• VS Code
• Azure Text Analytics (Language) resource

---

☁️ Azure Setup

Create:

• Resource Type: Text Analytics / Language
• Endpoint:

https://<your-resource>.cognitiveservices.azure.com/

• API Key

✅ Free or Standard tier both work

---

⚙️ Setup Instructions

1. Clone or Open Project

File → Open Folder → healthcare-nlp-app

---

2. Update docker-compose.yml

- eula=accept
- billing=https://YOUR-ENDPOINT.cognitiveservices.azure.com/
- apikey=YOUR_KEY
- rai_terms=accept

---

▶️ Run Application

Open terminal in VS Code:

docker compose down --volumes
docker compose up --build

---

🌐 Access

Service	URL
UI	http://localhost:8501
API	http://localhost:8000

---

🧪 Example Input

The patient was diagnosed with type 2 diabetes and prescribed metformin and insulin. He reports fatigue and thirst.

---

📊 Output

• Dashboard summary
• Diagnoses table
• Medications table
• Symptoms table
• Medical codes mapping
• JSON + downloadable report

---

🧬 Medical Coding Systems

System	Purpose
ICD-10	Billing
SNOMED CT	Clinical
UMLS	Mapping
DrugBank	Drugs

---

🌍 Access from Another Computer

1. Get IP:

ipconfig

2. Allow firewall:

netsh advfirewall firewall add rule name="Streamlit 8501" dir=in action=allow protocol=TCP localport=8501

3. Open:

http://YOUR-IP:8501

---

🛠️ Useful Commands

docker compose up --build
docker compose down
docker compose logs -f
docker ps

---

📦 Version

Current Version: v1.0.0

Initial release featuring:

• Clinical NLP extraction (Azure Text Analytics for Health)
• FastAPI backend
• Streamlit dashboard UI
• Dockerized deployment
• Structured clinical output with coding systems (ICD-10, SNOMED, UMLS, DrugBank)

---

📝 Changelog

v1.0.0 — Initial Release (March 26, 2026)

🚀 Features

• Extract diagnoses, medications, and symptoms from clinical text
• Map entities to ICD-10, SNOMED CT, UMLS, and DrugBank
• Interactive Streamlit dashboard with summary metrics
• FastAPI backend for processing requests
• Azure Text Analytics for Health integration
• Docker Compose multi-service architecture

🎨 UI Improvements

• Dashboard cards with icons
• Narrative summary generation
• Structured tables for extracted entities

📊 Output

• Clean JSON response
• Downloadable text report

---

🔮 Upcoming (Planned)

• PDF report generation
• Azure OpenAI clinical summary
• File upload (clinical notes / PDFs)
• Chat interface (RAG-based)
• Azure deployment (Container Apps)
• Authentication & multi-user support

🚀 Future Enhancements

• 📄 PDF Reports
• 🤖 AI Summary (LLM)
• 📁 File Upload
• ☁️ Azure Deployment
• 💬 Chat with Clinical Data

---

👤 Author

Sabin Shrestha
AI • Data Engineering • Healthcare Analytics

---

⭐ Why This Project Matters

👉 Converts messy clinical text → structured, billable, analyzable data

👉 Real-world use cases:

• EHR systems
• Insurance coding
• Clinical analytics
• AI pipelines

---

📌 Final Note

This is not just an NLP demo —
it is a production-ready healthcare AI system.