README.md

metadata

title: MRI Brain Tumor Detection
emoji: 🧠
colorFrom: blue
colorTo: purple
sdk: docker
app_port: 7860
pinned: false

🧠 MRI Brain Tumor Detection System

Deep Learning application for automated brain tumor classification from MRI scans using a custom ResidualInceptionBlock CNN architecture.

🎯 Features

• 4-Class Classification: Glioma, Meningioma, Pituitary, No Tumor
• Real-time Inference: Fast predictions with confidence scores
• Modern UI: Clean, responsive React interface
• RESTful API: FastAPI backend with automatic documentation

🏗️ Architecture

• Frontend: React 18 + Vite
• Backend: FastAPI + PyTorch
• Model: Custom ResidualInceptionBlock CNN (50+ layers)
• Deployment: Docker + Hugging Face Spaces

🚀 Quick Start

Using the Deployed App

Simply visit the app URL and upload an MRI scan image to get instant predictions.

Local Development

1. Clone the repository

https://huggingface.co/spaces/arghadip2002/NeuroGuard-Web-Application
cd NeuroGuard-Web-Application

2. Start with Docker Compose

docker-compose up --build

3. Access the application

• Frontend: http://localhost:3000
• API Docs: http://localhost:8000/docs

Manual Setup

Backend:

cd backend
pip install -r requirements.txt
uvicorn app.main:app --reload

Frontend:

cd frontend
npm install
npm run dev

📋 API Endpoints

• POST /api/predict - Upload MRI image for prediction
• GET /health - Health check endpoint
• GET /docs - Interactive API documentation

🎨 Usage

1. Upload an MRI brain scan (PNG, JPG, JPEG)
2. Click "Run Diagnosis"
3. View prediction with confidence score

📊 Model Information

• Classes: 4 (Glioma, Meningioma, Pituitary, No Tumor)
• Input Size: 224x224 RGB images
• Architecture: Custom ResidualInceptionBlock with 50+ layers

🛠️ Technology Stack

• PyTorch 2.1.0
• FastAPI 0.104.1
• React 18.2.0
• Vite 5.0.0
• Docker & Docker Compose

📝 License

MIT LICENSE

👨‍💻 Author

Arghadip Biswas and Sayan Das

📊 Dataset

• Dataset: https://github.com/arghadip2002/SAETCN-and-SASNET-Architectures/blob/main/dataLinks
• Based on SAETCN architecture

📜 Citation & Academic Acknowledgment

This repository provides the NeuroGuard Web Application, which is the deployment of the novel SAETCN and SAS-Net architectures detailed in our research paper.

If you use this deployed software or its code in your academic work, please cite the underlying paper to acknowledge the methodology and results:

✍️ Preferred Citation (BibTeX)

@article{das2025novel,
  title={Novel Deep Learning Architectures for Classification and Segmentation of Brain Tumors from MRI Images},
  author={Das, Sayan and Biswas, Arghadip},
  year={2025},
  eprint={2512.06531},
  archivePrefix={arXiv},
  primaryClass={cs.CV}
}

🔗 Paper Link

The full paper is publicly available on the arXiv preprint server: arXiv:2512.06531

⭐ Note: Citing the paper is essential for the advancement of open science and ensures proper credit for the research that powers this application. Thank you for your support!