SIC-Project---Brain-Tumor-Detection-using-VGG-16

🧠 Brain Tumor Detection Using VGG-16 Model

🎯 Project Overview

This project implements a Convolutional Neural Network (CNN) using the VGG-16 architecture for detecting brain tumors from MRI scans. The model achieves ~88% training accuracy and ~82% test accuracy using transfer learning techniques.

👨‍🎓 Project Information

• Author: Sohan Maity
• Roll No: 523EC0001
• Email: 523ec0001@iiitk.ac.in
• Institution: Indian Institute of Information Technology Design and Manufacturing Kurnool (IIITDM Kurnool)
• Program: Samsung Innovation Campus - Final Project

🎥 Project Presentation

📊 Dataset Information

The dataset contains MRI brain scan images classified into two categories:

• NO: No tumor detected (encoded as 0)
• YES: Tumor detected (encoded as 1)

Dataset Statistics

Set	Images	Accuracy
Training Set	169 images	~88%
Test Set	84 images	~82%

🏗️ Model Architecture

VGG-16 Transfer Learning

• Base Model: VGG-16 pre-trained on ImageNet
• Input Shape: (224, 224, 3)
• Frozen Layers: All VGG-16 convolutional layers
• Custom Head: Global Average Pooling + Dense layers

Model Structure

VGG-16 Base (Frozen)
    ↓
Global Average Pooling 2D
    ↓
Dense(1024, activation='relu')
    ↓
Dense(1024, activation='relu')
    ↓
Dense(512, activation='relu')
    ↓
Dense(2, activation='softmax')

🚀 Quick Start

Prerequisites

pip install tensorflow keras opencv-python matplotlib scikit-learn plotly tqdm imutils numpy

Installation

1. Clone the repository

   git clone https://github.com/sohan2311/brain-tumor-detection.git
   cd brain-tumor-detection

2. Install dependencies

   pip install -r requirements.txt

3. Download the dataset

   kaggle datasets download -d navoneel/brain-mri-images-for-brain-tumor-detection

Usage

1. Train the model

   python train_model.py

2. Make predictions

   python predict.py --image path/to/mri_scan.jpg

📁 Project Structure

brain-tumor-detection/
│
├── data/
│   ├── brain_tumor_dataset/
│   │   ├── yes/          # Tumor images
│   │   └── no/           # No tumor images
│   ├── TRAIN/
│   ├── TEST/
│   └── VAL/
│
├── models/
│   └── brain_tumor_model.h5
│
├── notebooks/
│   └── Brain_Tumor_Detection.ipynb
│
├── src/
│   ├── train_model.py
│   ├── predict.py
│   ├── data_preprocessing.py
│   └── model_architecture.py
│
├── requirements.txt
├── README.md
└── LICENSE

🔬 Technical Details

Data Preprocessing

• Image Resizing: 224×224 pixels
• Normalization: Pixel values scaled to [0, 1]
• Label Encoding: Binary classification (0: No Tumor, 1: Tumor)
• Data Split: 67% Training, 33% Testing

Model Configuration

• Optimizer: Adam (learning_rate=0.01)
• Loss Function: Categorical Crossentropy
• Metrics: Accuracy
• Epochs: 15
• Batch Size: Default

Performance Metrics

Metric	Training	Validation
Accuracy	88%	82%
Loss	0.0007	0.6297

📈 Results & Visualizations

Training History

The model shows excellent learning progression:

• Epoch 1: 62.47% → Epoch 15: 100% (Training Accuracy)
• Validation Accuracy: Stabilized around 89-90%

Key Findings

• ✅ Successful implementation of transfer learning
• ✅ Good generalization despite small dataset
• ✅ Effective feature extraction using VGG-16
• ⚠️ Signs of overfitting in later epochs

🧠 About Brain Tumors

A brain tumor is an abnormal mass of cells in the brain, which can be:

• Benign: Non-cancerous
• Malignant: Cancerous

Common Symptoms:

• Persistent headaches
• Seizures
• Vision/speech difficulties
• Memory loss
• Personality changes

Early detection through MRI analysis can significantly improve treatment outcomes.

🛠️ Technologies Used

📋 Requirements

tensorflow>=2.0.0
keras>=2.0.0
opencv-python>=4.0.0
matplotlib>=3.0.0
scikit-learn>=1.0.0
plotly>=5.0.0
tqdm>=4.0.0
imutils>=0.5.0
numpy>=1.19.0

🚀 Future Enhancements

• Improve Accuracy: Implement ResNet, EfficientNet architectures
• Data Augmentation: Add rotation, zoom, flipping techniques
• Multi-class Classification: Detect different tumor types (Glioma, Meningioma, Pituitary)
• Web Deployment: Create Flask/FastAPI web application
• Mobile App: Develop mobile application for real-time detection
• Model Optimization: Implement model quantization for faster inference

📊 Model Performance Visualization

The training process shows:

• Rapid Learning: Quick improvement in first 5 epochs
• Convergence: Model stabilizes around epoch 7-8
• Overfitting Signs: Training accuracy reaches 100% while validation plateaus

🤝 Contributing

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

1. Fork the project
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

🙏 Acknowledgments

• Samsung Innovation Campus for providing the learning platform
• IIITDM Kurnool for academic support
• Kaggle Community for the dataset
• TensorFlow/Keras Teams for the excellent deep learning frameworks
• VGG Team for the groundbreaking architecture

📞 Contact

Sohan Maity

• 📧 Email: 523ec0001@iiitk.ac.in
• 🎓 Institution: IIITDM Kurnool
• 📱 Roll No: 523EC0001

---

⭐ If you found this project helpful, please give it a star! ⭐

Made with ❤️ for advancing medical AI and brain tumor detection