- Data Exploration
- The Brain Tumor Detection 2020 Kaggle dataset, consisting of 3065 MRI images separated into two classes (Tumor โ Non-Tumor) brain, was used.
- Data Analysis
- cv2.COLOR_RGB2LAB was used to apply equalization histogram to perceptual lightness.
- Analysis of histograms indicated that the affected brains had much more perceptual lightness intensity.
- CNN Model
- The model included mainly 3 Conv2D layers with maxpooling and ReLU activation functions, followed by Flatten, Dense, Activation, Dropout, Dense and Activation layers.
- Model Training
- Binary crossentropy loss function was used for binary classification.
- Adam optimizer, a stochastic gradient descent optimization algorithm, was chosen to train the deep learning model as it can handle sparse gradients on noisy problems.
- Model Evaluation
- The model achieved a validation accuracy of 97% which indicates its potential for detecting brain tumors accurately.
- This project is Mathematical Modeling using PDEs for Detection of Brain Tumor which was required by our college to further join a competition where our team pirates got the 2nd place in it. The content of this project includes:
- Poster which illustrates our full project
- Implementing our algorithm using Deep Learning to detect brain tumor in Jupyter Notebook (Accuracy 97%)
- Report which describes the full research process
- Presentation to present our full idea
- NumPy
- Matplotlib
- Keras
- TensorFlow
- PIL
- SciPy
To run this project, install it locally using pip:
Shell
$ cd ../"project_path"
$ jupyter notebook brain-tumor-detection-notebook.ipynb
Name |
---|
Mahmoud Salman |
Ibrahim Mohamed |
Kamel Mohamed |
Youssef Shaaban |
Dina Khalid |
Youssef Kadry |
Neveen Mohamed |
Esraa Ali |