🧠 Brain Tumor Classification

Overview

This project focuses on classifying brain tumors using MRI images. It categorizes MRI scans into four distinct classes:

• Glioma
• Meningioma
• Pituitary
• No Tumor

We utilize the EfficientNetB1 model (pre-trained on ImageNet) and fine-tune it for this classification task. The dataset is sourced from Kaggle and contains labeled MRI scans for training and testing.

---

Key Features

• Data Preprocessing: Includes cropping MRI images to focus on the tumor area, resizing to a uniform size, and applying data augmentation techniques.
• Model Architecture: Utilizes EfficientNetB1 as a feature extractor followed by custom layers for classification.
• Training Enhancements:
  • Early Stopping
  • ReduceLROnPlateau for learning rate adjustment
  • Model Checkpointing to save the best model
• Visualization: Displays training/validation accuracy and loss graphs, confusion matrix, and classification reports for evaluation.

---

Installation & Setup

1. Clone Repository & Install Dependencies:

   !pip install opendatasets
   !pip install tensorflow
   !pip install imutils
   !pip install opencv-python
   !pip install matplotlib seaborn scikit-learn

2. Download Dataset:

   import opendatasets as od
   od.download("https://www.kaggle.com/datasets/masoudnickparvar/brain-tumor-mri-dataset/data")

---

Workflow Summary

1. Data Exploration & Visualization
2. Image Cropping Function: Focus on brain area by removing extra background.
3. Preprocessing & Saving Cropped Images for each class.
4. Image Augmentation using ImageDataGenerator (Rotation, Shift, Flip).
5. Model Building: Fine-tuning EfficientNetB1 with additional layers.
6. Training with Early Stopping and LR Scheduler.
7. Evaluation: Accuracy, Loss plots, Confusion Matrix, and Classification Report.

---

Model Performance

• Evaluation Metrics:
  • Training Accuracy & Loss
  • Validation Accuracy & Loss
  • Confusion Matrix
  • Classification Report

Graphs and metrics help in understanding the model's learning and generalization capabilities.

---

Google Colab 🚀

Highly Recommended:
For better GPU performance and faster training, it's advised to run the notebook on Google Colab.

You can upload your notebook to Google Colab and select GPU Runtime:

---

Folder Structure:

├── brain-tumor-mri-dataset/
│   ├── Training/
│   └── Testing/
├── Cropped_img/
│   ├── Train_Data/
│   └── Test_Data/
├── model.keras
├── README.md
└── brain_tumor_classification.ipynb

---

Dependencies

• TensorFlow
• OpenCV
• Matplotlib
• Seaborn
• Imutils
• Scikit-learn
• NumPy
• Pandas

---

Results

The model achieves excellent classification accuracy on both training and testing datasets, providing a robust solution for brain tumor detection from MRI images.

---

Future Improvements

• Implement transfer learning with deeper EfficientNet variants.
• Integrate attention mechanisms for better focus on tumor regions.
• Hyperparameter tuning for improved performance.
• Deployment of the model as a web application.

---