i wanna save my label tumor to nrrd, but why on the saving windows, i couldn't find segmentation-tumor.nrrd? i just found segmentation-label. any solution?
Hi Dennis, thanks for this useful tutorial. But I find certain aspects missing in my (latest to the date) version of the 3D slicer, such as Islands, crop volumes. Also, can you please guide on how to use Nvidia AIAA in 3D slicer.
Hi Talal. I just tried to update mine to the latest newest version (4.11.XXX) and indeed it seems like some menus have been rearranged. I found the crop volumes just by searching for it with the search function next to the menu panel. It is now found under converters -> crop volumes. The "islands" I am still able to find under the "segment editor" as shown in the video. I am running Max OSX. Maybe it looks different on another operating system? Unfortunately, I do not have any experience in using Nvidia AIAA in slicer3D.
Hi @Muhammad. Under file->DICOM (opens a DICOM browser), there you'll find an "import" button, to import your DICOM folder. You can also do as I show in the video, simple draw-n-drop your DICOM folder to the Slicer3D application.
Hi Dennis. Thank you for this videos. Given the number of manual segmentation one has to do to serve as ground truth. What is the practical number of sample data that can be used for training?
Hi Olukukola. That really depends on your domain. E.g. using 2D, 3D or some type of 4D images. The complexity of the structure, brain vs bone. As such, you should try to find a dataset that represents the distribution of structures. In some cases this can be done with a handful of images, for other structures, it requires thousands.
Hi Dennis ! Wonderful tutorial. Could you please help me with 3D tooth segmentation. I have 3D CT images of teeth and it is in the .STL file format. I'm struggling. No idea how to preprocess .stl files. Couldn't find any help from Google search. Anyone who is reading my comment please give me some insights on this. I would be very grateful.
Hi Anima, You do so with 3D slicer. When loading the data, just choose that it is a segmentation. Then in the segmentations module, you can export it to a labelmap.
Dennis, why there is no difference between the cortical bone and and the cancellous bone? I'd like CTs to be this way because they are easier to segment, is this because of the tomographer setup?
Hi Mario. For the application here, I just wanted to segment the full outer surface of the femural bone. I actually never specifically looked at the cancellous bone. It seems that it has a slightly darker intensity. So it should be possible to just set a higher threshold level and thereby segment the cancellous bone. Some manual cleanup might also be needed though as it will probably have similar intensity as the tissue around the femural bone as well.
@@DennisMadsen I confused the bone marrow with the trabecular bone. How do you ensure that the CT does not differentiate the bone from the bone marrow? I want that because it is easier to segment the outer surface of the bone. Is it because a configuration of the tomographer?
Hi Mario. Are you referring to the acquisition of the images? I do not know how to configure the CT scanner for image acquisition. I am only working with the images that I am provided with. In the case shown - a CT image of a leg.
@@DennisMadsen Yes, I'm referring to the acquisition of the images. Usually an axial slice of a leg looks like this: ibb.co/bdsQyDb which is different from what is seen in the video time 3:41... that's why I ask
Thanks a lot for the clarification. I do not know the exact reason for this difference. I would assume that either the two images are taken with different detail level (slicing distance), or the image contrast as I show in Slicer might be different than the one you link to (this can be changed as shown at 3:22). I have honestly never looked at this, as I am mainly looking at the full bone structures to perform shape analysis on.
Essentially you can make any kind of segmentation in slicer3d. If none of the semi-automatic methods works, you can still manually annotate either multiple slides at once, or annotate the image volume slice by slice.
Thanks for your guidance and effort, Dennis. I can't wait for your next video!
Amazing video!! thanks i am looking forward for the second and next parts
Glad you liked it!
Thanks a lot.. helped me a lot as a beginner
Thank you Dennis. This was helpful. I would also like to know how to save the segmented labels in such a way that I can edit then later.
I finally found out that the NRRD file has the 3D segmented labels. Thanks
Glad to hear that you found a solution :)
i wanna save my label tumor to nrrd, but why on the saving windows, i couldn't find segmentation-tumor.nrrd? i just found segmentation-label. any solution?
Hi Tsania, Have a look around 9:28 when I export the labelmap. In the export panel you have to export labelmap. Hope this helps.
Hi Dennis, thanks for this useful tutorial. But I find certain aspects missing in my (latest to the date) version of the 3D slicer, such as Islands, crop volumes.
Also, can you please guide on how to use Nvidia AIAA in 3D slicer.
Hi Talal. I just tried to update mine to the latest newest version (4.11.XXX) and indeed it seems like some menus have been rearranged. I found the crop volumes just by searching for it with the search function next to the menu panel.
It is now found under converters -> crop volumes.
The "islands" I am still able to find under the "segment editor" as shown in the video. I am running Max OSX. Maybe it looks different on another operating system?
Unfortunately, I do not have any experience in using Nvidia AIAA in slicer3D.
Hi. The dataset that I have is in .dcm format. It has slices of each 3D representation. Any idea on how should I visualize them in 3D?
Hi @Muhammad. Under file->DICOM (opens a DICOM browser), there you'll find an "import" button, to import your DICOM folder. You can also do as I show in the video, simple draw-n-drop your DICOM folder to the Slicer3D application.
Hi Dennis. Thank you for this videos. Given the number of manual segmentation one has to do to serve as ground truth. What is the practical number of sample data that can be used for training?
Hi Olukukola. That really depends on your domain. E.g. using 2D, 3D or some type of 4D images. The complexity of the structure, brain vs bone. As such, you should try to find a dataset that represents the distribution of structures. In some cases this can be done with a handful of images, for other structures, it requires thousands.
Alright. Many thanks
Hi Dennis ! Wonderful tutorial.
Could you please help me with 3D tooth segmentation. I have 3D CT images of teeth and it is in the .STL file format. I'm struggling. No idea how to preprocess .stl files. Couldn't find any help from Google search.
Anyone who is reading my comment please give me some insights on this.
I would be very grateful.
Hi Anima,
You do so with 3D slicer. When loading the data, just choose that it is a segmentation. Then in the segmentations module, you can export it to a labelmap.
Dennis, why there is no difference between the cortical bone and and the cancellous bone? I'd like CTs to be this way because they are easier to segment, is this because of the tomographer setup?
Hi Mario. For the application here, I just wanted to segment the full outer surface of the femural bone. I actually never specifically looked at the cancellous bone. It seems that it has a slightly darker intensity. So it should be possible to just set a higher threshold level and thereby segment the cancellous bone. Some manual cleanup might also be needed though as it will probably have similar intensity as the tissue around the femural bone as well.
@@DennisMadsen I confused the bone marrow with the trabecular bone. How do you ensure that the CT does not differentiate the bone from the bone marrow? I want that because it is easier to segment the outer surface of the bone. Is it because a configuration of the tomographer?
Hi Mario. Are you referring to the acquisition of the images? I do not know how to configure the CT scanner for image acquisition. I am only working with the images that I am provided with. In the case shown - a CT image of a leg.
@@DennisMadsen Yes, I'm referring to the acquisition of the images. Usually an axial slice of a leg looks like this: ibb.co/bdsQyDb which is different from what is seen in the video time 3:41... that's why I ask
Thanks a lot for the clarification. I do not know the exact reason for this difference. I would assume that either the two images are taken with different detail level (slicing distance), or the image contrast as I show in Slicer might be different than the one you link to (this can be changed as shown at 3:22). I have honestly never looked at this, as I am mainly looking at the full bone structures to perform shape analysis on.
thanks for the video brother
Is this AI enhanced ! Or manual in 3d slicer
Can we do brain tumor segmentation on slicer3d
Essentially you can make any kind of segmentation in slicer3d. If none of the semi-automatic methods works, you can still manually annotate either multiple slides at once, or annotate the image volume slice by slice.
Hi Dennis, is there a way to contact you?
Hi @Reenergise, I've now updated the about section. Send me a PM on twitter or LinkedIn :-)