This project is a Streamlit-based AI-powered diagnostic assistant designed to support healthcare professionals in remote or underserved areas. By leveraging multiple intelligent agents and Google’s Gemini language model, this tool provides an AI-driven differential diagnosis based on patient data such as symptoms, vitals, lab results, and medical history.
- Intuitive Web Interface built with Streamlit.
- Multi-Agent Workflow using CrewAI for specialized medical data analysis.
- Natural Language Diagnosis powered by Google's Gemini model (via
litellm). - Structured Diagnostic Summary provided in both on-screen format and downloadable markdown.
- Clear, human-readable explanation of the diagnostic reasoning.
- Modular agents for:
- Symptom analysis
- Vital signs assessment
- Lab report interpretation
- Medical history evaluation
- Follow-up recommendations
- User Input: Clinician inputs symptoms, vitals, lab results, and patient history.
- CrewAI Agents: The app organizes and preprocesses the data using a structured team of specialized agents.
- Diagnosis Synthesis: Agents collaboratively generate a comprehensive diagnosis and suggested follow-up.
- User Output: A structured diagnostic summary and explanation are displayed and can be downloaded.
Each agent has a clear medical role (e.g., Symptom Interpreter, Vital Sign Analyzer) and contributes to a shared diagnostic objective using a shared memory pipeline orchestrated by CrewAI.
- Streamlit – for UI development
- CrewAI – to simulate a team of expert AI agents
- LiteLLM – lightweight interface to use Gemini and other LLM APIs
- Google Gemini API – for large language model inference
- Python – core language for logic, orchestration, and data handling
- 🧪 Add support for uploading detailed lab reports and medical images.
- 📊 Visualize diagnosis trends and graphs for better interpretation.
- 🔐 Add user authentication and patient data security layers.
- 🌍 Integrate offline capability or lightweight model for truly remote deployments.
- 🌐 Add multilingual support for better global usability.
- 📱 Package as a mobile-friendly PWA (Progressive Web App).
- Python 3.8+
- pip
- A valid Google Gemini API key
-
Clone the Repository
git clone https://github.com/SrujanPR/AI-Diagnostic-Assistant.git cd ai-diagnostic-assistant -
Create a Virtual Environment (optional but recommended)
python -m venv venv source venv\Scripts\activate
-
Install Dependencies
pip install -r requirements.txt
-
Add Your API Key
Replace the API key in
app.py(GeminiLLM class) with your own Google AI Studio API key. -
Run the App
streamlit run app.py
-
Sample Input
- Symptoms: swelling in legs, fatigue, reduced urine output, nausea
- Body Temperature: 98.7
- Blood Pressure: 145/95
- Pulse Rate: 88
- Lab Results: Creatinine 3.5, BUN 52, Hb 9.8
- History: diabetes, hypertension
When you enter these inputs and run the program one of the probable diagnosis should be Chronic Kidney Disease (CKD) as these are the symptoms of CKD.
ai-diagnostic-assistant/
│
├── app.py # Main Streamlit app with CrewAI logic
├── requirements.txt # Python dependencies
├── README.md # This file
Built by SRUJAN P R
Feel free to reach out for collaborations, ideas, or improvements.
Pull requests are welcome! If you’d like to improve the assistant or contribute new agents or features, feel free to fork the repo and submit a PR.
This project is licensed under the MIT License. See the LICENSE file for more details.
This tool is not a substitute for professional medical advice. It is designed to assist, not replace, qualified healthcare providers. Always consult a physician for serious or urgent medical concerns.