🧠 Brain Disease Segmentation Using Deep Learning

A comprehensive deep learning project for detecting and segmenting brain diseases, particularly tumors, in MRI scans using multiple state-of-the-art architectures including U-Net and Meta's Segment Anything Model (SAM).

🌟 Key Features

• Multi-Model Support: Implementations using both U-Net and SAM architectures
• Advanced Segmentation: Pixel-level precise segmentation of brain anomalies
• Automated Detection: Accurate identification of brain tumors and diseases
• Interactive Visualization: Rich set of visualization and analysis tools
• High Performance: GPU-optimized implementation with excellent metrics
• Flexible Pipeline: Support for various data formats and preprocessing techniques

🏗️ System Architecture

U-Net Implementation

graph TD
    subgraph Data Pipeline
        A[Raw MRI Scans] --> B[Image Processing]
        B --> C[Dataset Creation]
        C --> D[Data Augmentation]
    end
    
    subgraph U-Net Architecture
        E[Input Layer] --> F[Encoding Path]
        F --> G[Bottleneck]
        G --> H[Decoding Path]
        H --> I[Output Layer]
        F -.-> H
    end

Loading

SAM Implementation

graph TD
    A[Input MRI Image] --> B[SAM Processor]
    B --> C[Vision Encoder]
    B --> D[Prompt Encoder]
    C --> E[Mask Decoder]
    D --> E
    E --> F[Segmentation Mask]
    E --> G[Probability Map]

Loading

📊 Performance Metrics

• DICE Coefficient: 85-87%
• IoU Score: 78%
• Precision: 89%
• Recall: 83%

🛠️ Technical Stack

• Deep Learning Frameworks: PyTorch, HuggingFace Transformers
• Segmentation Models: U-Net, SAM (Segment Anything Model)
• Image Processing: OpenCV, Albumentations
• Analysis Tools: Pandas, NumPy, MONAI
• Visualization: Matplotlib
• Development: Python 3.x

🚀 Getting Started

Installation

# Clone the repository
git clone https://github.com/AmirrHussain/segmentation-of-brain-diseases.git

# Install requirements
pip install -r requirements.txt

# Install SAM
pip install git+https://github.com/facebookresearch/segment-anything.git

# Additional dependencies
pip install -q git+https://github.com/huggingface/transformers.git
pip install -q monai

Directory Structure

├── main.py           # Main execution file
├── config.py         # Configuration settings
├── brain.py         # Neural network architectures
├── trainer.py       # Training functions
├── plotter.py       # Visualization tools
├── sam_utils.py     # SAM-specific utilities
└── dataSets/        # MRI scan datasets

Dataset Organization

dataSets/
├── patient_1/
│   ├── scan.jpg
│   └── scan_mask.jpg
└── patient_2/
    ├── scan.jpg
    └── scan_mask.jpg

💻 Usage Examples

U-Net Training

python main.py --model unet --mode train

SAM Inference

from transformers import SamModel, SamProcessor

processor = SamProcessor.from_pretrained("facebook/sam-vit-base")
model = SamModel.from_pretrained("facebook/sam-vit-base")
outputs = model(**processor(image, input_boxes=[[prompt]], return_tensors="pt"))

🔄 Future Roadmap

• Multi-modal MRI support
• Web-based interface
• Enhanced augmentation pipeline
• Model ensemble implementation
• 3D volume segmentation
• Real-time processing capabilities

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request:

1. Fork the repository
2. Create your feature branch (git checkout -b feature/AmazingFeature)
3. Commit changes (git commit -m 'Add AmazingFeature')
4. Push to branch (git push origin feature/AmazingFeature)
5. Open a Pull Request

📄 License

This project is licensed under the Apache 2.0 License - see the LICENSE file for details.

📬 Contact

Hossein Karimi

• Email: hk.karimi.93@gmail.com
• GitHub: @AmirrHussain

🔗 Important Links

• Project Repository
• LGG Segmentation Dataset
• SAM Documentation

✨ Acknowledgments

• Dataset providers
• PyTorch and HuggingFace communities
• Medical imaging experts
• Meta AI Research team