MRI scans and patient data are sensitive. Ensure proper anonymization and adhere to ethical guidelines when working with medical datasets.

Duration: Dec 2020 to May 2021

This project employs a Convolutional Neural Network (CNN) to detect brain tumors from MRI scans. Using Python libraries like TensorFlow, Keras, and OpenCV, the model identifies whether an MRI scan contains a brain tumor with high accuracy.

Brain-Tumor-Detection/
├── brain_tumor_detection.ipynb  
├── README.md
└── requirements.txt                

1. Data Preprocessing:

   • MRI scan images are resized, normalized, and converted to grayscale to prepare them for model training.

   • Labels are assigned based on the presence or absence of a tumor.

2. Model Architecture:

   • A Convolutional Neural Network (CNN) is designed using Keras with TensorFlow as the backend.

   • The network consists of convolutional layers for feature extraction and fully connected layers for classification.

3. Training:

   • The model is trained on the labeled dataset using supervised learning.

   • Cross-entropy loss is minimized using the Adam optimizer.

4. Detection:

   • The trained model predicts whether an MRI scan contains a tumor.

   • The results are evaluated using metrics such as accuracy, precision, and recall.

• Supervised Learning:The project uses labeled MRI data to train the CNN. Each image is paired with a label indicating whether it contains a tumor.

• Convolutional Neural Networks:CNNs are ideal for image classification tasks. They extract spatial hierarchies of features through layers of convolutions, pooling, and activation functions.

• Metrics:

  • Training Accuracy: 99%

  • Validation Accuracy: Between 89% and 92%

  • Evaluated using a confusion matrix and loss plots.

1. Clone the repository

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

2. Install dependencies

   pip install -r requirements.txt

3. Run the Jupyter Notebook:

   Open brain_tumor_detection.ipynb in Jupyter Notebook or Google Colab.

4. Prepare data:

   • Place MRI scan images in the data directory.

   • Update file paths in the notebook if necessary.

5. Train or use the pretrained model:

   • To train from scratch, follow instructions in the notebook.

   • To use the pretrained model, load trained_model.h5.

• Training Loss and Accuracy:Loss and accuracy plots over epochs are generated to visualize model performance.

• Prediction Example:Sample predictions are displayed in the notebook, including tumor detection results on test images.

• Python: Primary language for implementing machine learning models and data analysis.

• TensorFlow & Keras: Built and trained the Convolutional Neural Network.
• OpenCV: Preprocessed MRI images for resizing and grayscale conversion.
• NumPy: Handled numerical operations like array transformations.
• Pandas: Managed and manipulated dataframes for easier data handling.
• Matplotlib: Visualized training metrics (loss and accuracy plots).
• Scikit-learn: Evaluated model performance using metrics like confusion matrix and classification report.

Feel free to reach out if you have questions or suggestions:

• Email: [email protected]

• GitHub: ashish-shiju