The goal of this project is to develop a Convolutional Neural Network (CNN) model to classify MRI scans as either containing a tumor or not. The model utilizes both CNN and Deep Neural Network (DNN) for binary classification, with accuracy serving as the primary metric to assess its performance.
Data set Description:
The images in this project use the Kaggle data set:Brain MRI Images for Brain Tumor Detection。
Two categories in total::
- NO - no brain tumor
- Yes - has brain tumor
Note: The author of this data set did not indicate the source of the brain tumor images.
View the project code : BrainMRI
| software | version |
|---|---|
| Python | 3.11.5 |
| Matplotlib | 3.7.2 |
| Numpy | 1.25.2 |
| Pandas | 2.0.3 |
| Tensorflow | 2.12.1 |
| keras | 2.12.0 |
| Topic | Explanation |
|---|---|
| kaggle data set | Brain MRI Images for Brain Tumor Detection |
| Data set explanation | 253 files in total, including 98 files for no and 155 files for Yes. |
| Reference | Brain MRI |
This project combines a CNN model classification problem (used to predict whether a subject has a brain tumor) and a computer vision problem (used to automate the process of cropping brains from MRI scans). The final accuracy is much higher than the baseline of 50% (random guessing). However, this accuracy can be further improved by increasing the number of training images or by adjusting model hyperparameters.