I found sub-OAS30001_ses-d0757_T2w more suitable for tinkering. And it’s really that, because these studies target the brain and not blood vessels.
Without contrast, it is difficult to identify veins and arteries in slice views, at bifurcations namely.
You may use the ‘Grow from seeds’ effect of the ‘Segment editor’. Place seeds with 2 segments like below in a few slices where you can identify the structures of interest.
The green is the target structure, the yellow for anything else. Preview, fix with more seeds and apply. I could thus get the result below. Arteries and veins are fused. Each seeding will produce its unique result.
You may also try ‘Guided vein segmentation’ in SlicerVMTK extension. Place an open curve following what you identify as the carotid artery, set parameters and apply. I could get the result below. Better, but not much extensible towards the the skull.
Overall, figuratively, it’s spending time trying to create a valuable byproduct from waste material. Sometimes we have good results, sometimes not. It’s better to use series that study the carotids specifically, if possible.
Thanks, this has been really helpful. I am just having a bit of trouble knowing exactly where to place the seeds. When I placed the seeds I got nothing like what you got for the 3d. Maybe can you show me where you placed the seeds in the image? and also how do you use the guided vein segmentation. It always asks me for an open curve and I have know idea what that is. Can you please explain to me what that is?
Well I might not be understanding your question in full.
You want to segment the carotids. Identify these while scrolling in slice views. Every time you think a structure is the carotid, place a seed there with the ‘Paint’ effect of the segment editor. And all around it, a few seeds of another segment that target anything around. The first image is explicit.
The seeding step is totally on you, as well as refining the preview with more seeds.
This is one method where you do everything by yourself. If you go for the second method referenced above, it’s globally less demanding and the result may be more satisfactory.
Yeah my question was kinda phrased bad. I am really new to medical imaging and so I have no idea if I am being honest where the carotid arteries are in the image, so I was thinking if you could like export the “seeds” drawing (segmentation) so I then can just download that and then use the grow from seeds to replicate what you did in the second image.
Here is an MRB file with the seeds. It contains the initial seeds and those that refine the preview.
Remember that every seeding gives a unique result, so the final result is different from that in the second image above, better in this case.
You should probably do some training with datasets that study blood vessels. In the sample module of Slicer, CTA-cardio is a good starting point. Learning segmentation from inadequate volumes will be much deceiving, it’s like trying to run before learning to walk.
Thanks this has been really helpful. Is this all the segmentation you did, because when I download your .mrb I only get a little of bit what you got for the second image
It’s just the seeding that you requested. You must then select ‘Grow from seeds’ effect in the ‘Segment editor’ module, preview, fix with more seeds as required and apply.
You’ll get more encouraging results if you play with contrast enhanced CT scans.
No, applying contrast is a physical event. It is injected in the subject’s vein by the assistant of the radiologist. It’s loaded in the injector device that is remotely controlled. That holds true for both MRI and CT scans, though different products.
Hi. I’m back once again. I am trying to segment the carotid arteries in an MRA (Magnetic Resonance Angiogram), and I have about 900 images in the dataset. I was using the threshold effect from SegmentEditorExtraEffect, but it isn’t accurate, and every time, I have to use the scissors function and delete some of the unnecessary junk from the threshold. And this take me about thirty minutes to do for each image, and so to do it for about 900 images, it would take ages, and I am kinda of in a time cramp right now, so I was wondering if there are other ways to segment the carotid arteries in an MRA much faster than the way I am doing right now. Maybe there is an extension on github I can download and then upload to 3d slicer and use that? If anyone wants to view my data I will provide a link to some of it: OASIS-3 Angiogram Dataset - Google Drive
Using sub-OAS30001_ses-d2430_acq-TOF_angio_1, I could get all contrasted arteries with the ‘Threshold’ effect of the ‘Segment editor’, and it’s really fast. The noise can easily be removed using the ‘Islands’ effect with ‘Keep selected island’. The online manual will help you concerning how these effects should be used.
Hi. I have another quick question. I am trying to save both the original image and the segmentation into one file. How is that possible? The only way I am seeing is mrb, but I need either a Dicom file format or NIFTI. So if you know any way to do this please help me out!
