This project is focused on the detection of brain tumors from MRI scans using deep learning. It originated as a final project and has since been updated and expanded. The project explores two different models for cancer detection: one for detecting meningioma tumors and another for classifying multiple types of brain tumors.

• Meningioma Tumor Detection: A model to classify MRI scans as either positive or negative for meningioma tumors, achieving an accuracy of 93%.
• Multi-Class Tumor Detection: A more advanced model that detects and classifies three types of brain tumors (Glioma, Meningioma, and Pituitary), achieving an accuracy of 95%.
• Transfer Learning: Utilizes the InceptionV3 architecture with weights pre-trained on ImageNet for the multi-class detector, demonstrating the effectiveness of transfer learning for medical imaging tasks.
• Jupyter Notebooks: The repository includes the Jupyter notebooks used for training, evaluation, and prediction, providing a clear view of the entire workflow.

This project includes two distinct models for brain tumor detection.

This project is focused on the detection of brain tumors from MRI scans using deep learning. It originated as a final project and has since been updated and expanded. The project explores two different models for cancer detection: one for detecting meningioma tumors and another for classifying multiple types of brain tumors.

• Meningioma Tumor Detection: A model to classify MRI scans as either positive or negative for meningioma tumors, achieving an accuracy of 93%.
• Multi-Class Tumor Detection: A more advanced model that detects and classifies three types of brain tumors (Glioma, Meningioma, and Pituitary), achieving an accuracy of 95%.
• Transfer Learning: Utilizes the InceptionV3 architecture with weights pre-trained on ImageNet for the multi-class detector, demonstrating the effectiveness of transfer learning for medical imaging tasks.
• Jupyter Notebooks: The repository includes the Jupyter notebooks used for training, evaluation, and prediction, providing a clear view of the entire workflow.

This project includes two distinct models for brain tumor detection.

This model is a Convolutional Neural Network (CNN) built from scratch using TensorFlow. It is designed for the binary classification of meningioma tumors.

• Architecture:
  • The model consists of three convolutional layers with 16, 32, and 64 filters, respectively.
  • The ELU (Exponential Linear Unit) activation function is used in all convolutional and dense layers.
  • MaxPooling2D is applied after each convolutional layer to downsample the feature maps.
  • Two fully-connected (Dense) layers with 128 units each follow the convolutional layers.
  • The final output layer uses a softmax activation function for classification.
• Regularization: To prevent overfitting, the model employs two regularization techniques:
  • Dropout with a rate of 0.2 is applied after each pooling layer and between the dense layers.
  • L2 regularization is applied to the weights of all convolutional and dense layers.
• Performance: This model achieves an accuracy of 93% on the test set.

This model is designed to classify MRI scans into four categories: Glioma, Meningioma, Pituitary tumor, or no tumor. It leverages transfer learning to achieve high accuracy.

• Technique: Transfer learning and fine-tuning.
• Base Model: The InceptionV3 model, pre-trained on the ImageNet dataset, is used as the base for feature extraction. The original classification head of InceptionV3 is removed.
• Custom Head: A new classification head is added on top of the InceptionV3 base, which includes:
  • A GlobalAveragePooling2D layer.
  • A Dense output layer with 4 units and a softmax activation function.
• Training Process:
  1. Feature Extraction: The model is first trained with the InceptionV3 base frozen (weights are not updated). This allows the new classification head to adapt to the brain tumor dataset.
  2. Fine-Tuning: After the initial training, the InceptionV3 base is unfrozen, and the entire model is trained with a very low learning rate. This fine-tunes the pre-trained weights to be more specific to the task of tumor detection. During this phase, the BatchNormalization layers in InceptionV3 are kept frozen to stabilize training.
• Data Augmentation: To improve generalization, the training data is augmented with RandomFlip (horizontal) and RandomRotation.
• Performance: This model achieves an accuracy of 95% on the test set.

The models were trained on datasets from Kaggle:

• Brain Tumor Classification (MRI): https://www.kaggle.com/datasets/sartajbhuvaji/brain-tumor-classification-mri
• Brain Tumor MRI Dataset: https://www.kaggle.com/datasets/masoudnickparvar/brain-tumor-mri-dataset

To use this project, you will need to have Python, TensorFlow, and other standard data science libraries installed. The Transfer.ipynb and Train&Test.ipynb notebooks in the Cancer_Detector and Meningioma_Detector directories, respectively, contain the code for training the models.

The pre-trained model file for the multi-class detector can be found at the following link: https://drive.google.com/drive/folders/1o7ts623pJQxxuOs5kQBkyjEyorH8lT0X?usp=sharing