Brain Tumour Detector

Application of Machine Learning, AI and Data Mining methods, such as YOLOv8 model and Convolutional Neural Networks (CNNs) for building a model capable of detecting tumours in brain CT scans.

• We used a public dataset available in Kaggle to develop the project. It's publicly available at the following link: Medical Image DataSet: Brain Tumor Detection;

• We built the project based on an existing Jupyter notebook, also publicly available at Kaggle: Brain Tumor Detection w/Keras YOLO V8;

• Although the achieved results weren't satisfactory, we constructed a model which CIoU was almost 3x lower than the original model and a mAP almost 7x higher. The optimization tweaks also heavily reduced the training time (more than 6x faster);

• If you want to see the deployed application, click down below and feel free to test the models with your own instances and visualize a static dashboard about the dataset:

  • Deploy:

1. Requirements

• Python3 and pip package manager:

  sudo apt install python3 python3-pip build-essential python3-dev
  

• virtualenv tool:

  pip install virtualenv
  

• Libraries:

  • Machine Learning and Data Mining: Keras, KerasTuner, TensorFlow, imbalanced-learn;
  • Computer Vision: OpenCV, KerasCV;
  • Data Analysis, Visualization and Manipulation: pandas, Streamlit, Plotly express, Kaleido, seaborn, Matplotlib, numpy;
  • Others: PyCryptodome, Pillow, gdown and google-colab.

• Environments: Jupyter.

2. Web App

In this section, you can see the the detector GUI made with Streamlit.

3. Execution

In this section, you can follow detailed instructions for executing the project.

1. Clone the repository

   git clone https://github.com/juliorodrigues07/tumour_detection.git
   

2. Enter the repository's directory

   cd tumour_detection
   

3. Create a virtual environment

   python3 -m venv .venv
   

4. Activate the virtual environment

   source .venv/bin/activate
   

5. Install the dependencies

   pip install -r requirements.txt
   

3.1. Detector and Dashboard

• You first need to be in the src directory to run the command:

  streamlit run 1_🏠_Home.py
  

3.2. Data Mining

• To visualize the notebooks online and run them (Google Colaboratory), click on the following links:

  • EDA;
  • Data Mining.

• To run the notebooks locally, run the commands in the notebooks directory following the template: jupyter notebook <file_name>.ipynb.

  • EDA (Exploratory Data Analysis):

    jupyter notebook 1_eda.ipynb
    

  • Data Mining:

    jupyter notebook brain_tumor_detection_w_keras_yolo_v8.ipynb
    

4. Project Structure

.
├── README.md                             <- Project's documentation
├── requirements.txt                      <- File containing all the required dependencies to run the project
├── plots                                 # Directory containing all the graph plots generated
├── assets                                # Directory containing images used in README.md and in the deployed app
├── datasets                              # Directory containing all used or generated datasets in the project
|   ├── image_statistics.csv              <- Statistical data about the dataset (std, mean, channels, ...)
|   ├── labels.csv                        <- Tumour types and quantities data
|   └── coords.csv                        <- Detections data (coordinates and area)
├── docs                                  # Directory containing all the presentation slides about the project      
|   ├── Transparencies - Partial I.pdf
|   ├── Transparencies - Partial II.pdf
|   └── Transparencies - Final.pdf          
└── models                                # Directory containing all generated models in the project
|   ├── base.keras                        <- Trained with vanilla dataset
|   ├── reduced.keras                     <- Trained with reduced dataset
|   └── balanced.keras                    <- Trained with balanced dataset
├── notebooks                             # Directory containing project's main jupyter notebook
|   ├── 1_eda.ipynb
|   └── brain_tumor_detection_w_keras_yolo_v8.ipynb
└── src                                   # Directory containing the web application
    ├── 1_🏠_Home.py                      <- Main page with the tumour detector
    └── pages                             # Child pages directory
        └── 2_📊_Static.py                <- Script responsible for generating the static dashboard

5. Outro

• To uninstall all dependencies, run the following command:

  pip uninstall -r requirements.txt -y
  

• To deactivate the virtual environment, run the following command:

  deactivate
  

• To delete the virtual environment, run the following command:

  rm -rf .venv