MediNodus (Latin for 'Healing Knot') untangles the complexity of medical data. It is an intelligent mobile platform that translates complex lab reports into simple language and prevents dangerous drug interactions using real-time FDA data.
Medical literacy is a massive barrier to healthcare in India:
- Complex Reports: Patients struggle to understand lab results (e.g., "What does high Lymphocytes mean?").
- Adverse Interactions: Elderly patients often take multiple prescriptions without knowing if they conflict, leading to preventable hospitalizations.
MediNodus acts as a bridge between the doctor's data and the patient's understanding.
- 📄 Report Translator: Upload a PDF/Image of a blood test. The AI extracts values, flags abnormalities, and explains them in plain English (or Hindi).
- 💊 Drug Safety Shield: Scan a medicine strip. The app identifies the drug and cross-references it with your current health profile to warn of risks (e.g., "Don't take Ibuprofen; your Kidney function is low").
- 🧠 MedGamma Intelligence: Uses a fine-tuned local Transformer model to understand medical text and imagery without external APIs.
- 🛡️ Privacy First: No data processing on third-party clouds (Google/OpenAI). All inference happens within your container.
The system follows a microservices-ready architecture, containerized with Docker for instant scalability.
(Insert your Mermaid Architecture Diagram here)
| Component | Technology | Role |
|---|---|---|
| Mobile App | React Native (Expo) | Cross-platform UI for scanning and history. |
| Backend API | FastAPI (Python 3.10) | Async API handling inference requests. |
| AI Engine | Hugging Face Transformers | Running the local MedGamma model. |
| Database | MongoDB Atlas (Beanie) | Storing user profiles and structured history. |
| Media Storage | Cloudinary | Hosting encrypted medical documents. |
- On-Device/On-Prem Inference: Uses the
transformerslibrary to run MedGamma. - Abnormality Detection: Parses lab values and flags "Critical" statuses.
- Latency Optimized: No network calls to LLM providers.
- Interaction Checker: Checks Drug-Drug interactions using local knowledge bases.
- Personalized Warnings: Detects conflicts between a new drug and your existing Chronic Conditions.
- History Tab: View past scans with visual thumbnails.
- Secure Archive: Reports are indexed and retrievable.
- Node.js & npm
- Python 3.10+
- Docker & Docker Compose
- MongoDB Atlas URI
git clone https://github.com/yourusername/medinodus.git
cd medinodus
Create a .env file in the backend/ directory:
MONGO_URL=mongodb+srv://<user>:<password>@cluster.mongodb.net/?retryWrites=true&w=majority
CLOUDINARY_KEY=<your_cloudinary_secret>
SECRET_KEY=<your_jwt_secret>
HF_TOKEN=<hf_auth_token>
# No External AI Keys Required!
Configure API URL: Create a new file named .env inside the mobile/ folder and add your backend URL:
# For Local Testing (Replace with your actual PC IP, not localhost)
EXPO_PUBLIC_API_URL=[http://192.168.1.5:8000](http://192.168.1.5:8000)The Docker build will install PyTorch and Transformers for local inference.
cd backend
docker build -t medinodus-backend .
docker run -p 8000:8000 --env-file .env medinodus-backend
Note: The first run may take time to download the MedGamma model weights to the container.
cd mobile
npm install
npx expo start
Scan the QR code with the Expo Go app on Android/iOS.
- Gaurav Upreti: Android/React Native Engineering
- Juss PAtel: Backend Systems & Database Architecture
- Ashmeet Singh Sandhu: AI/ML Pipelines & Research
Distributed under the MIT License. See LICENSE for more information.