BRATS-Image-Segmentation

This repository provides a comprehensive pipeline for preprocessing 3D medical images, particularly from the BraTS2019 dataset. These preprocessing steps are designed to prepare images for deep learning models in medical imaging tasks.

Preprocessing Steps

• N4 Bias Field Correction: Corrects intensity non-uniformity in MRI images.
• Cropping: Removes empty regions, focusing on the brain area.
• Image and Mask Preprocessing: Normalizes images and prepares masks, saving them as X.npy and Y.npy.
• Dataset Creation: Filters slices with no tumor regions to balance the dataset.
• Processing Folds: Divides data into 5 folds for cross-validation.
• Data Structuring: Combines processed images and masks for model training.
• Data Splitting: Splits data into training and validation sets. Available here

Repo's directory structure

The directory structure below shows the nature of files/directories used in this repo.

BRATS-Image-Segmentation
│
├── README.md
│
├── data              
│   ├── BraTS2019              # Main data
│   ├── interim                # Intermediate data that has been processed
│   └── processed
│       ├── train              # Processed training data
│       └── valid              # Processed validation data
│
├── models
│   └── Model.png              # Model architecture visualization
│
├── notebooks
│   ├── Brats_DL.ipynb         # Notebook for deep learning model
│   └── Brats_Prep.ipynb       # Notebook for preprocessing
│
└── src
    ├── preprocessing
    │   ├── N4biasfieldcor.py  # Bias field correction
    │   ├── image_cropper.py   # Crop 3D images
    │   ├── image_processor.py # Process 3D images and masks
    │   ├── create_folds.py    # Data creation and folding
    │   ├── utils.py           # Preprocessing utilities
    │   └── __init__.py
    ├── Modules.py             # DLUNetModel and ImageVisualizer classes
    ├── __init__.py
    └── utils.py               # General utilities