Segmentation of a binary image

[Wo0p]

Hello PF users,

This is my first time on these forums. I'm not sure this is exactly the right place for this question, but as I couldn't find an image processing sub forum I'm posting it here in the hopes someone can help me!

Homework Statement

I have a binary image (see below) derived from a grayscale MRI image by thresholding.

The contrast I'm using is supposed be good for differentiating WM/GM. The problem is that it also picks up a lot of skull and noise (you can see it on the edge of the brain) which are lumped together with WM. I want to get rid of this non-brain tissue and create a WM-only mask.

Homework Equations

N/A

The Attempt at a Solution

What I've done before is to use ImageJ to manually remove the obviously non-brain voxels. This is very traumatizing and depressing. How would you solve this problem?

 

[LightHero]

Is there a simple algorithm I can implement where I define a boundary for the brain and it automatically removes all non-brain tissue? Thank you in advance!

 

[Mene]

Segmentation of a binary image is the process of dividing the image into distinct regions or segments based on certain criteria or features. In your case, you are trying to segment the image to separate the white matter (WM) from other non-brain tissue and noise.

One approach you can try is to use a morphological operation called "opening" on the binary image. This operation involves eroding the image followed by dilating it, which can help remove small objects and smooth out the edges. You can adjust the size of the structuring element used in the operation to control the amount of noise removal.

Another approach is to use a region-growing algorithm, where you select a seed point within the WM region and grow the region by iteratively adding neighboring pixels that meet certain criteria (e.g. similar intensity values). This can help differentiate the WM from other tissues and noise.

You can also explore other segmentation techniques such as thresholding based on multiple intensity levels or using machine learning algorithms.

Overall, the best approach will depend on the specific characteristics of your image and the level of accuracy required for your segmentation. It may also be helpful to consult with a colleague or seek guidance from an image processing expert for further assistance.