MRI TUMOR CLASSIFICATION

📌 Overview

This project implements a deep learning-based MRI tumor classification model using PyTorch. The model is trained on MRI brain scans to classify different tumor types using a CNN architecture.

🚀 Features

• Preprocessing: Image transformations, data augmentation, and dataset splitting.
• Deep Learning Model: Custom Convolutional Neural Network (CNN) for MRI classification.
• Training & Evaluation: Model training, validation, and evaluation metrics.
• Visualization: Data distribution, confusion matrix, class performance, and Grad-CAM heatmaps.
• Testing & Prediction: Evaluate model accuracy and classify new MRI images.

---

📂 Project Structure

├── data_preparation.py   # Prepares dataset, applies transformations
├── model_definition.py   # Defines CNN model
├── train_and_evaluate.py # Training & evaluation logic
├── test_accuracy.py      # Calculates test set accuracy
├── testSample.py         # Predicts MRI tumor class for a given image
├── main.py               # Orchestrates data processing, training & testing
├── visualization.py      # Contains visualization functions
├── models/               # Stores trained model weights
├── plots/                # Stores various plots (confusion matrix, ROC curves, etc.)
└── metrics/              # Stores model performance metrics

---

🔧 Installation & Setup

1️⃣ Clone the Repository

git clone https://github.com/mertmetin1/MRI-Tumor-Classification.git
cd MRI-Tumor-Classification

2️⃣ Install Dependencies

pip install -r requirements.txt

3️⃣ Dataset Preparation

Ensure the dataset is structured as follows:

./data/Raw/
 ├── Training/
 │   ├── glioma/
 │   ├── meningioma/
 │   ├── notumor/
 │   ├── pituitary/
 ├── Testing/
 │   ├── glioma/
 │   ├── meningioma/
 │   ├── notumor/
 │   ├── pituitary/

---

🏋️‍♂️ Training the Model

Run the main.py script to train the model:

python main.py

This will:

• Load the dataset
• Train the CNN model
• Save the best-performing model
• Generate evaluation plots

---

🧪 Testing & Prediction

Test Model Accuracy

python test_accuracy.py

Classify a New MRI Image

python testSample.py --image_path path/to/image.jpg

---

📊 Visualization & Evaluation

• Confusion Matrix (plots/confusion_matrix.png)

• 

• ROC Curves (plots/roc_curve.png)

• 

• Class-wise Performance (plots/per_class_accuracy.png)

• 

• Grad-CAM Heatmaps (plots/grad_cam/)

• 

---

📈 Model Performance

Metric	Value
Best Validation Accuracy	97.46%
Final Training Loss	0.0272
Final Validation Loss	0.1888
Test Accuracy	98.63%

---

🤖 Model Architecture

The CNN model consists of 3 convolutional layers, ReLU activations, MaxPooling, and fully connected layers:

Conv2D(3 → 32) → ReLU → MaxPool
Conv2D(32 → 64) → ReLU → MaxPool
Conv2D(64 → 128) → ReLU → MaxPool
Flatten → Fully Connected → Output (Softmax)

For a detailed visualization, see plots/model_architecture.png.

---

📝 License

This project is licensed under the MIT License.

---

🙌 Acknowledgments

Special thanks to the open-source community for providing tools like PyTorch & TorchVision for deep learning research.

---

🔗 References

• PyTorch Documentation
• Torchvision Transforms