Cancer Detection Web Application 🔬

A deep learning-powered web application for detecting cancer in medical images. Built with Streamlit and PyTorch, this application uses a fine-tuned ResNet34 model to provide real-time cancer detection from medical imaging data.

Features

• Real-time cancer detection from medical images
• Interactive web interface built with Streamlit
• Probability distribution visualization with confidence scores
• Support for TIF format medical images
• System information monitoring
• Data augmentation during training
• Transfer learning using pretrained ResNet34

Model Architecture

The application uses a modified ResNet34 architecture:

• Base: Pretrained ResNet34 from torchvision
• Modifications:
  • Frozen feature extraction layers
  • Custom fully connected layer (num_features → 2 classes)
  • Optimized for binary classification

Training Details

• Dataset Split:
  • Training: 2608 images
  • Validation: 100 images
  • Test: Remaining images
• Data Augmentation:
  • Random horizontal flips
  • Random rotation (±5 degrees)
  • Resize to 224x224
  • Normalization (ImageNet statistics)
• Training Parameters:
  • Optimizer: Adam (lr=3e-4)
  • Loss Function: CrossEntropyLoss
  • Batch Size: 32
  • Workers: 2
• Hardware: Supports both CPU and GPU training

Requirements

python >= 3.6
streamlit
torch
torchvision
Pillow
huggingface_hub
pandas
numpy
tqdm

Installation

1. Clone the repository:

git clone https://github.com/yourusername/cancer-detection-webapp.git
cd cancer-detection-webapp

2. Create a virtual environment (recommended):

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install the required packages:

pip install -r requirements.txt

4. Configure Hugging Face credentials:
   • Create an account on Hugging Face
   • Replace YOUR_HUGGING_FACE_USERNAME with your username in app.py
   • For private repositories, set your HF token:

     export HF_TOKEN=your_token_here  # On Windows: set HF_TOKEN=your_token_here

Usage

Training the Model

1. Prepare your dataset:

   data/
   ├── data_sample/
   │   ├── image1.tif
   │   ├── image2.tif
   │   └── ...
   └── labels.csv
   

2. Run the training script:

python train_model.py

Running the Web Application

1. Start the Streamlit application:

streamlit run app.py

2. Open your web browser and navigate to http://localhost:8501

3. Use the application:

   • Upload a medical image (.tif format)
   • Click "Analyze Image"
   • View results and probability distribution

Model Performance

The model is evaluated on a held-out test set with metrics including:

• Binary classification accuracy
• Loss convergence during training
• Real-time performance metrics available in the web interface

Contributing

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

Development Roadmap

Future improvements planned:

• Support for additional image formats
• Model explainability visualizations
• Batch processing capabilities
• GPU acceleration support
• REST API endpoint
• Docker containerization

License

MIT

Acknowledgments

• ResNet34 architecture from torchvision
• Web interface built with Streamlit
• Model hosting on Hugging Face Hub

Screenshots

Citation

If you use this project in your research or work, please cite it as:

@software{cancer_detection_webapp,
  author = {[Faraz Shamim]},
  title = {Cancer Detection Web Application},
  year = {2024},
  url = {https://github.com/mr-imperium/cancer-detection}
}