AyuUnity

A full-stack health-tech web application that combines medical-image analysis, chronic-disease risk prediction, symptom checking, and an LLM-backed assistant in a single dashboard. Built with Next.js (frontend) and Flask (backend).

What it does

The Flask backend exposes three core ML endpoints:

Endpoint	Task	Model
POST /analyze/<scan_type>	Medical image classification (e.g. brain tumor, skin lesion)	PyTorch CNN — BrainTumorCNN (4 classes), SkinLesionCNN (2 classes)
POST /predict-chronic-risk	Diabetes + multi-disease chronic-risk scoring from a form input	scikit-learn classifiers (chronic_model.pkl, chronic_multi_model.pkl)
POST /check-symptoms	Disease prediction from free-text symptoms	TF-IDF + Random Forest (s2d_*.pkl), with Gemini for follow-up suggestions

Gemini is used for medical suggestions and disease info on top of the model outputs.

Frontend pages

• / — landing
• /login, /register-patient
• /patient-dashboard, /doctor-dashboard — role-specific dashboards
• /scan-analysis — image upload + result display
• Components for chronic-risk form, symptom checker, prescription list, government schemes, video consultation, emergency button, chatbot

Tech Stack

• Frontend: Next.js, React, TypeScript, Tailwind CSS, shadcn/ui (Radix primitives)
• Backend: Flask, Flask-CORS
• ML: PyTorch (image classification), scikit-learn (tabular + text), TF-IDF
• NLP: NLTK preprocessing (stopwords, lemmatizer)
• LLM: Google Gemini (google-generativeai) for medical suggestions

Project Structure

AyuUnity/
├── backend/
│   ├── app.py                      # Flask routes for all 3 endpoints
│   ├── utils/
│   │   ├── model_loader.py         # PyTorch CNNs + sklearn loaders
│   │   └── gemini_integration.py   # Gemini wrapper
│   ├── models/                     # Trained .pth and .pkl models
│   ├── uploads/                    # Runtime upload folder
│   ├── requirements.txt
│   └── .env.example
└── frontend/
    ├── app/                        # Next.js App Router pages
    ├── components/                 # Domain components + shadcn UI
    ├── public/, styles/, lib/, hooks/
    └── package.json

Setup

Backend

cd backend
python -m venv venv
source venv/bin/activate          # Windows: venv\Scripts\activate
pip install -r requirements.txt

cp .env.example .env
# Set GEMINI_API_KEY

python app.py

Backend runs at http://localhost:5000.

Frontend

cd frontend
npm install                       # or pnpm install
npm run dev

Frontend runs at http://localhost:3000.

Models

The backend/models/ directory holds the trained model files used at runtime:

File	Purpose
brain_tumor_model.pth	PyTorch CNN — 4-class brain MRI classification
skin_lesions_model.pth	PyTorch CNN — binary skin lesion classification
chronic_model.pkl	scikit-learn — diabetes risk
chronic_multi_model.pkl	scikit-learn — multi-condition chronic risk
s2d_Random_model.pkl	Random Forest — symptom-to-disease
s2d_tfidf_vectorizer.pkl	TF-IDF vectorizer for symptom text
s2d_label_encoder.pkl	Label encoder for disease names

Notes

This was built as a hackathon-style prototype. It's a working full-stack demo, not a regulatory-grade clinical tool — none of the models are FDA/CE certified and outputs should not be used for diagnosis.