README.md

AI Medical Research Assistant Prototype

⚠️ This is a research prototype only - NOT FOR MEDICAL USE ⚠️

Project Description

A proof-of-concept AI system demonstrating:

• Medical image analysis (X-rays, CT scans)
• Prescription text parsing
• Research-backed medical Q&A using DeepSeek
• Safety guardrails for medical content

Features

• 📷 Basic medical image analysis
• 💊 Prescription medication extraction
• 🔍 Research-supported responses
• ⚠️ Automatic safety filtering

Setup Instructions

1. Hugging Face Secrets:

   • Add your DeepSeek API key as a secret:
     • Name: DEEPSEEK_API_KEY
     • Value: [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]

2. Required Models:

   • Will auto-download on first run:
     • obi/deid_bert_i2b2 (Prescription parser)
     • Hate-speech-CNERG/dehatebert-mono-english (Safety filter)
     • TorchXRayVision models

3. Hardware Requirements:

   • GPU Space recommended
   • Minimum 8GB RAM

Usage

1. Input options:

   • Medical question (text)
   • Medical images (X-ray/CT scans)
   • Prescription text

2. Output:

   • Research-supported response with safety disclaimers

Models Used

Component	Model/API	Source
Medical Q&A	DeepSeek API	https://deepseek.com
Prescription Parsing	obi/deid_bert_i2b2	Hugging Face Model Hub
X-ray Analysis	DenseNet121	TorchXRayVision
Safety Filter	dehatebert-mono-english	Hugging Face Model Hub

Legal Disclaimer

THIS SYSTEM:

• ❌ Is NOT a medical device
• ❌ Cannot diagnose conditions
• ❌ Should NOT be used for treatment decisions
• ✅ For research purposes only

Consult licensed healthcare professionals for medical advice.

Limitations

• Image analysis uses placeholder implementations
• Prescription parsing limited to English text
• Responses require clinical validation
• No HIPAA compliance implemented

---

Warning: This prototype contains simulated medical AI components. Actual medical AI systems require regulatory approval and clinical validation.