MRI Brain Tumor Segmentation

Hello everyone.

I'm working on an image preprocessing project for school. I need to segment an MRI brain image to detect a tumor. I'm using the BraTS dataset.

I've tried various denoising techniques (median filter, Non-Local Means filter, Gaussian filter, Anisotropic diffusion filter), as well as different contrast enhancement techniques (CLAHE, histogram equalization, imadjust) and segmentation methods (Otsu, thresholding, k-means). However, none of these approaches have given me satisfactory results. I only get a stylized version of the brain.

Can anyone help me? What techniques should I use, and how would you implement them?

Thank you very much for your help!

Answer 1

[Gayathri]

Hi @D,

For effective segmentation, it is better to use Deep Learning techniques.

Please refer to the below link, which will give you an overall idea of "3-D Brain Tumor Segmentation Using Deep Learning".

https://www.mathworks.com/help/images/segment-3d-brain-tumor-using-deep-learning.html

This example uses a 3-D pretrained U-Net for segmentation. You can also try using different models for effective segmentation.

One such model is "DeepLab v3+". It uses atrous (or dilated) convolutions, which allow the network to capture multi-scale contextual information without losing resolution. This is particularly useful for segmenting objects at various scales.

For more information on how to train a segmentation model using “deeplabv3plus” refer to the following link.

https://www.mathworks.com/help/vision/ref/deeplabv3plus.html

[D]

I know that Deep Learning techniques are better, but unfortunatly I can not use them

[Walter Roberson]

Sometimes the correct answer is, "None of these techniques work very well."