Brain-Tumor-Segmentation-using-CNN-in-MRI-images.

To develop a Convolutional Neural Network (CNN) segmentation model to diagnose brain tumor using Magnetic Resonance Imaging (MRI) images.
The project was run on Google Colab that provided a single 12GB NVIDIA Tesla K80 GPU.

   These are 155 slices per MRI sequence of a particular patient.  

Dependencies :

1. Numpy
2. Matplotlib
3. Keras
4. SimpleITK

Dataset :

For downloading the dataset

1. Go to "https://www.smir.ch/BRATS/Start2015".
2. Register in there with official E-mail id.
3. After confirming they will send the logins to your E-mail id.
4. Login in there and go to "Challenges/BRATS2015".
5. Then download the training and train dataset. But they won't provide the ground truth for test dataset.

Files Description

In the codes folder following files are there :

1. utils.py : It is the utility script containing the data loading and data augmentation code.
2. simple_model.py : It is the utility script containing simple CNN model implemented in the paper.
3. unet_model.py : It is the utility script containing the Unet model.
4. Brain_segmentation.ipynb : It is notebook containing both the models executed 5 and 35 epochs respectively.
5. Assets : It contains the sample dataset for just one patient and some files linked to notebook

Contributors

• Rahul Kumar

License & copyright

© Rahul Kumar 2020
Licensed under the MIT License