Tumor Segmentation in MRI Brain Scans

This project performs image segmentation on MRI brain image to identify and segment tumor regions using binary masking and labeling techniques in Python.

---

📌 Table of Contents

• Project Overview
• Dataset
• Objective
• Methodology
• Results
• How to Run
• Challenges
• Contributors

---

📖 Project Overview

In this project, I used binary masking technique to segment tumors in MRI brain scan slices. The segmented images highlight tumor boundaries, and distinguish multi-region tumors using intensity differences and region labeling.

This can assist medical practitioners or serve as a proof-of-concept for more advanced models.

---

Dataset

• Type: MRI scan brain images

• Format: .nii (Neuroimaging Informatics Technology Initiative format)

• Slices: 2D grayscale slices from 3D volumes
  📥 Download Link

• Full Data Source

• 📥 Download Link

---

Objective

• Detect and visualize tumor regions in brain CT images.
• Label nested and overlapping tumors with separate colors.
• Improve visibility by applying masks and filtering borders.
• Enable easy inspection and interpretation of segmented regions.

---

Methodology

1. Data
   • Loaded .nii image using nibabel
2. Segmentation via binary masking
   • Tumor regions identified by pixel intensities (image > 8000) & (image <= 10500) pixel values
   • Separated high and low intensity tumors using color-mesh
3. Masking
   • Binary masks created using NumPy
4. Border Clearing
   • skimage.segmentation.clear_border() used to remove artifacts
5. Labeling
   • skimage.measure.regionprops() to identify tumor regions and their properties such as area, bounding boxes, center_of_mass
   • scipy.measure.label() to label connected regions and identified region areas
6. Visualization
   • matplotlib.pyplot.pcolormesh() for color visualization
   • scipy.ndimage.zoom() to reduce axes dimensions
   • plotly.graph_objects() for 3D visualisation of segmented tumors

---

📊 Results

Example segmented slice (slice index = 110):

Original	Segmented

Nested tumors:

• Small tumor : Higher intensity
• Larger tumor: Moderately Lower intensity, re-labeled as a distinct region

---

▶ How to Run

git clone https://github.com/Amos77Robert/segmentation-of-brain-tumors-without-AI.git
cd segmentation-of-brain-tumors-without-A

## 🧪 Dependencies
- numpy
- matplotlib
- nibabel
- scipy
- scikit-image

Install them via:
```bash
pip install -r requirements.txt

# Run the notebook in VS Code, Colab, Jupyter or any Python environment
python iMAGE sEGMENTATION sCRIPT.ipynb

---

Challenges

• Differentiating nested tumors with close intensity ranges
• Choosing optimal thresholds across variable slices

---

👤 Contributors

• Amoss Robert
  • Master of Data Science Candidate | Researcher