/brain-tumor-classification-cnn

Brain Tumor Detection and Classification using CNN

Primary LanguageJupyter Notebook

Brain Tumor Detection and Classification with Convolutional Neural Networks

Kaggle notebook link - https://www.kaggle.com/code/cheesecke/brain-tumor-classification-using-cnn/

Dataset Used - https://www.kaggle.com/datasets/masoudnickparvar/brain-tumor-mri-dataset

Introduction

This project utilizes TensorFlow and Keras to build and train a CNN model for brain tumor classification. The dataset used for training and evaluation contains MRI images of brain tumors, with each image labeled according to its tumor type.

Model Architecture

The CNN model architecture used for brain tumor classification consists of multiple convolutional layers followed by fully connected dense layers. Here's a detailed explanation of the model architecture:

  1. Input Layer:

    • Input images are expected to have a size of 224x224 pixels with 3 color channels (RGB).

  2. Convolutional Layers:

    • The model starts with a series of convolutional layers (Conv2D) with 64 filters, each having a kernel size of (3, 3), using the ReLU activation function.
    • This pattern repeats, increasing the number of filters.
    • Each convolutional layer uses "same" padding to ensure the output feature maps have the same spatial dimensions as the input.
    • After every two convolutional layers, a max-pooling layer (MaxPooling2D) with a pool size of (2, 2) is added to reduce spatial dimensions and extract dominant features.

  3. Flatten Layer:

    • The output feature maps from the last convolutional layer are flattened into a one-dimensional array.

  4. Dense Layers:

    • A series of fully connected dense layers (Dense) follows the flatten layer.
    • The first dense layer has 256 units with ReLU activation, followed by another dense layer with 64 units and ReLU activation.
    • The final dense layer has a number of units equal to the total number of classes in the dataset (determined by class_count) and uses the softmax activation function to output class probabilities.

  5. Compilation:

    • The model is compiled using the Adamax optimizer with a learning rate of 0.001.
    • Categorical cross-entropy is used as the loss function since this is a multi-class classification problem.
    • Accuracy is chosen as the evaluation metric.

Results

The trained CNN model achieved promising results in classifying brain tumor images. The evaluation metrics, including accuracy, precision, recall, and F1-score, demonstrate the model's effectiveness in accurately identifying different tumor types.

Confusion Matrix

Contributing

Contributions are welcome! If you'd like to contribute to this project, please follow these steps:

  1. Fork the repository.
  2. Create a new branch (git checkout -b feature/improvement).
  3. Make your changes and commit them (git commit -am 'Add new feature').
  4. Push the changes to your fork (git push origin feature/improvement).
  5. Create a new Pull Request.