marcinbrzezanski/brain-tumor-mri-classification

Project made in Jupyter Notebook with Kaggle Brain tumors 256x256 dataset, which aims at the classification of brain MRI images into four categories, using custom CNN model, transfer learning VGG16 model and fine tuning.

Brain tumor MRI classification

Table of Contents

• Introduction
• Data Description
• Methodology
• Technology Stack

Introduction

Project made in Jupyter Notebook with Kaggle Brain tumors 256x256 dataset, which aims at the classification of brain MRI images into four categories, using different CNN models.

Data Description

Dataset contains 3096 images of human brain MRI scans divided into four categories:

• Pituitary tumor
• Meningioma tumor
• Giloma tumor
• Normal - no tumor

Methodology

1. Data importing & augmentation - Notebook starts with importing data and applying rescaling, rotation and horizontal flip to perform data augmentation.
2. Data Visualization - To get better understanding of data, sample from every category is being displayed along with it's label.
3. Custom CNN model - Custom Convultional Neural Network model is being build, trained and evaluated by accuracy score and Loss & Accuracy charts.
4. Transfer Learning - With help of VGG16 pretrained convultional base another model is developed, trained and evaluated as earlier.
5. Fine tuning VGG16 model - By allowing some of VGG16's convultional layers to be trainable model is being further trained to fit data better.

Technology Stack

• Python 3.9.18
• Pandas 2.0.3
• Numpy 1.26.0
• Matplotlib 3.8.0
• Tensorflow 2.6.0
• Keras 2.6.0