This dataset contains 7023 human brain MRI images classified into 4 classes: glioma - meningioma - no tumor and pituitary. no tumor class images were taken from the Br35H dataset, while SARTAJ dataset has a problem that glioma class images are not categorized correctly.
First: Image Processing
1. Cropping to focus on brain shape
2. Applying Gaussian Filter for blurring on gray scale images
3. Apply Thresholding
4. apply active contour for tumor shape extraction
Second: Feature Extraction
1. Mean, Standard Deviation, Variance, Skewness, Entropy, RMS, Kurtios, HMI1, HMI2, HMI3, HMI4 of image
2. Area, Perimeter, Circular shape, Convex Area, Solidity, Equivalent Diameter, Major Axis, Minor Axis, of The tumor after image contouring
3. GLCM features like contrast, homogeneity, energy, correlation, dissimilarity
Third: Modeling
1. Random Forest Classifier (90%)
2. XGBoosting Classifier (91%)
3. Support Vection Machine (88%)
4. Knearest Knieghbour (87%)
Fourth: Hyperparameter tuning and Feature selection
* Scaling data using Standard Scaler (increased the test accuracies of models)
* Dimensionality Reduction Using PCA (was not a good option)
* best feature was
Fifth: Evaluation
* Confusion Matrix
* F1 Score, Precision, and Recall
main
├─ Notebooks
│ ├─ Feature Extraction
│ ├─ Modelling and Evaluation
│ └─ Data Analysis
├─ Paper & contributions
├─ Contour Images & processed images
└─ README.md
Download it from www.python.org/downloads/
- Install the following packages
pip install numpy
pip install Flask
pip install os
pip install PIL
pip install opencv-python
pip install skimage
Eights Semester - Biomedical Digital Signal Processing (SBE3110) class project created by:
Team Members' Names | Section | B.N. |
---|---|---|
Ahmed Hassan | 1 | 4 |
Habiba Fathalla | 1 | 27 |
Rahma Abdelkader | 1 | 31 |
Yousr Ashraf | 2 | 54 |
- Dr. Enas & Eng. Merna All rights reserved © 2023 to Team 6 - Systems & Biomedical Engineering, Cairo University (Class 2024)