Medical Document Processor

A FastAPI application that processes medical documents using Large Language Models (LLMs), extracts structured data, implements Retrieval-Augmented Generation (RAG), and converts data to FHIR format.

Features

• Document Management: Store and retrieve medical documents with SQLAlchemy
• LLM Integration: Summarize medical notes and extract patient information using OpenAI
• RAG Pipeline: Question answering system using vector embeddings and ChromaDB
• Structured Data Extraction: Extract patient data, diagnoses, medications with ICD/RxNorm codes
• FHIR Conversion: Convert structured data to FHIR-compatible format
• Containerized Deployment: Full Docker support for easy deployment

Architecture

This application demonstrates a complete medical document processing pipeline:

1. FastAPI Backend - RESTful API with automatic documentation
2. LLM Integration - OpenAI GPT models for text processing
3. RAG System - Vector-based document retrieval and question answering
4. Medical Code Lookup - Integration with public health APIs (ICD-10, RxNorm)
5. FHIR Compliance - Healthcare data format standardization
6. Docker Containerization - Easy setup for testing

Prerequisites

• Docker and Docker Compose
• OpenAI API key

Quick Start

1. Clone the Repository

git clone https://github.com/mjv57/medical-doc-processor.git
cd medical-doc-processor

2. Set Up Environment Variables

Create a .env file in the project root:

Option 1: Using command line

cat > .env << EOF
OPENAI_API_KEY=your-openai-api-key-here
APP_ENV=production
LOG_LEVEL=info
EOF

**Option 2: Manually create with below:

OPENAI_API_KEY=your-openai-api-key-here APP_ENV=production LOG_LEVEL=info

3. Build and Start the Application

A. Build the Docker image

docker build -t medical-document-processor .

B. Start the application

docker-compose up -d

C. View logs

docker-compose logs -f

4. Access the Application

Once running, the API will be available at:

• Interactive Documentation: http://localhost:8000/docs
• Health Check: http://localhost:8000/health

5. Manual Testing - Curl commands

Health Check

curl http://localhost:8000/health

Summarize a Document

curl -X POST "http://localhost:8000/documents/1/summarize" \
  -H "Content-Type: application/json" \
  -d '{"document_id": 1, "use_cache": true}'

Ask a Question (RAG)

curl -X POST "http://localhost:8000/answer_question" \
  -H "Content-Type: application/json" \
  -d '{"text": "What was the patient blood pressure?"}'

Extract Structured Data

curl -X POST "http://localhost:8000/documents/1/extract_structured"

Convert to FHIR Format

curl -X POST "http://localhost:8000/documents/1/to_fhir"

6. Python Test Files

Test the agent service functionality

python test_agent.py

Test FHIR conversion capabilities

python test_fhir.py

Test LLM API integration

python test_llm_api.py

Test RAG (Retrieval-Augmented Generation) functionality

python test_rag.py

7. Stopping the Application

docker-compose down -v

docker image rm medical-document-processor

Troubleshooting

# View application logs
docker-compose logs -f

docker-compose logs app

System Compatibility

Getting this error when running docker-compose up can result in a failure to start the application.

The requested image's platform (linux/amd64) does not match the detected host platform (linux/arm64/v8) and no specific platform was requested.

IF the application does not start, implement Fix: add below line above build in docker-compose.yml

services:
  app:
    platform: linux/arm64  # Fix here
    build: .
    container_name: medical-document-processor

Rebuild following fix:

docker-compose down -v
docker build --no-cache -t medical-document-processor .
docker-compose up -d