Fracture segmentation in CT scanning images

segmentation

#1

Recently I received a bunch of Tiff image data of CT scanning. I want to analyze the tiny fractures on the cross sections. But since the fracture on the top right part of the images was situated in some black stuff. It’s really hard to get the real fracture out of it for the little width of the fracture. Do you have a good idea dealing with it? Thank you very much!

You can find the attached link to the images!
https://drive.google.com/drive/folders/1-uWEz1cQ4d_2hxBgKcodv1CorotgOKKg?usp=sharing


#2

Goopd day,

does this


help?

Regards

Herbie


#3

Is this the Enhance Contrast- Equalize histogram? I guess? But after this operation, you can’t segment the fracture pretty well by thresholding?


#4

But after this operation, you can’t segment the fracture pretty well by thresholding?

You didn’t tell us what you mean by “segment the fracture”.
This is much too general. You need to tell what you like to measure in the end.

In any case, the main problem is noise and you won’t get rid of it without loosing detail.
Best is to spend more X-ray photons to get better image quality, especially less noise.

Regards

Herbie


#5

My original motivation is to segment the fractures and try to measure the width, length and orientation. So it would be great if I could segment the fractures as precise as I can. So you thought the quality is the main problem to achieve it?


#6

So you thought the quality is the main problem to achieve it?

That’s what I wrote …

measure the width, length and orientation

You won’t get much better images as this:
Fractures

You must decide whether this is good enough for what you want. For reliable automatic measurements it is not suited.
But why not do it manually?

Good luck

Herbie


#7

Hi Herbie. Really appreciate your timely response. Manual segmentation is good choice, but I have 800 images. The fractures have some variation for different images. So how do you achieve the images above and how do you think of using Trainable weka segmentation on your images?

Sincerely,
Charles


#8

Charles!

Manual segmentation is good choice, but I have 800 images.

The earlier you start the sooner you are finished!
How comes that you need to do this analyses without prior tests?

how do you think of using Trainable weka segmentation on your images?

I wouldn’t even try!

The image quality is insufficient for what you like to automatically obtain.
Noise reduction leads to blur and loss of detail and as you can see the fractures appear discontinous and patchy.

The image posted before was lowpass-filtered with a Gaussian of sigma = 2. Then a hand-set gray-value windowing and contrast inversion was applied, nothing that can be sufficiently automated.

It simply doesn’t look promising for reasonable automated analyses.

Regards

Herbie


#9

Hi Herbie:

Since I am a newcomer to this software. I didn’t expect much information from these CT data until I found Image J. I thought maybe I could do something.

The first thing you did was applying Gaussian Blur. . .(Filter), Right? Then how do you adjust the gray value and contrast simultaneously?
The last thing is about manual segmentation. You mean I should use freehand selection tools to outline them?

Best regards,
Charles


#10

Charles,

as a newcomer and now with such a mighty tool as ImageJ and such images it doesn’t surprise that you are lost.

I didn’t expect much information from these CT data […]

and you are right!

I’m tempted to say, forget about image processing and ask youself what would you do without. And why didn’t you try with some samples first before 800 images have been acquired.

I really don’t understand your role in the whole process. Did you promise someone that you can do the job without knowing how?

The last thing is about manual segmentation. You mean I should use freehand selection tools to outline them?

I don’t know.
What was your initial idea?

Clueless

Herbie


#11

@Charles_hoo

If you want to brush up on Segmentation in ImageJ/Fiji - here are some helpful links to get you started:

eta


#12

Hi Herbie:

CT scanning is just part of my project. I plan to continue to do some SEM scanning in detail. With CT scanning images, I could find some ROI and put it to SEM. Lately, I get to know this software and want to get some information for reference.

Best,
Charles


#13

Thank you very much for your guidance, I will check them out carefully!

Best,
Charles