A full-stack, AI-powered healthcare platform bridging the gap between immediate skin lesion screening and professional clinical care. > Developed for the CSI Project Expo by Team 27.
Skin cancer is the most common cancer globally, but when detected early, the 5-year survival rate for melanoma is a staggering 99%. Unfortunately, wait times for specialists can stretch for months, and generic internet searches only cause unnecessary panic.
AI Dermatologist is an accessible, AI-powered preliminary screening platform. We didn't just build a basic scanner; our custom Convolutional Neural Network (CNN) is trained on over 3,000 verified clinical patient images, achieving an impressive 97% accuracy in predicting whether a lesion is benign or malignant. Rather than leaving users stranded with a scary result, our platform seamlessly connects them to real medical professionals by generating secure clinical reports and instantly locating nearby hospitals.
- 🤖 AI Image Analysis: Upload a photo of a skin lesion for instant, clinical-grade classification (Benign/Malignant) powered by our custom TensorFlow model, boasting a 97% detection accuracy.
- 🗺️ Interactive Body Mapping: Users can visually pinpoint the exact location of the mole on a body map to build a personalized risk profile.
- 📄 Secure PDF Reports: Automatically generates downloadable, professional medical reports containing AI confidence metrics and patient history.
- 🔐 Hospital Portal & Secure Share Code: Each scan generates a unique 6-digit code. Doctors can log into the dedicated Hospital Portal, enter this code, and instantly import the patient's AI report and details before the consultation even begins.
- 🏥 Real-Time Clinical Connection: Integrated Leaflet maps utilize the Overpass API to instantly locate dermatologists and clinics within a 5-kilometer radius.
- React.js & TypeScript: Core UI framework.
- Vite: Lightning-fast frontend build tool.
- Tailwind CSS & shadcn/ui: Responsive, modern styling.
- Framer Motion: Smooth UI animations.
- Python & Flask: Lightweight, high-performance RESTful API.
- Flask-CORS: Secure cross-origin communication.
- Overpass API: Real-time geographic data for mapping hospitals.
- TensorFlow & Keras: Custom CNN architecture for image classification.
- NumPy & Pillow (PIL): Image processing and matrix transformations.
- Google Colab: Cloud GPU training environment used for the 3,000+ images.
- Leaflet.js: Interactive geographic mapping.
- jsPDF: Secure, client-side report generation.
- Web Storage API: Local patient database simulation.
- Vercel: Serverless frontend hosting.
Follow these instructions to run the full-stack application on your local machine. You will need two terminal windows running simultaneously.
- Node.js installed
- Python 3.8+ installed
git clone [https://github.com/CSI-Project-Expo/Team-27-.git](https://github.com/CSI-Project-Expo/Team-27-.git)
### 2. Start the Python AI Backend
Open your first terminal and navigate to the backend directory.
Bash
cd backend
# Install required Python packages
pip install -r requirements.txt
# Run the Flask server
python app.py
### 3. Start the React Frontend
Open a new, second terminal window and navigate to the frontend directory.
Bash
cd frontend
# Install dependencies
npm install
# Start the development server
npm run dev