Hi Folks,

I was trying to use LongCruise pipeline to extract the brain surfaces, but it turned out that in its very first step, SPECTRE failed to extract the brain correctly; some part of the brain is missing. Then I tried my own another brain extraction method and segmented the extracted image using TOADS, it seemed that TOADS had problem too. All I used were default settings. Could anybody tell me what's wrong?

The images are attached.

Thanks,
Sophia

Hi,

It is not clear from these images how the original image looks in the area that is not segmented correctly by TOADS, an axial view of GM membership and hard segmentation might be useful. So what structure is that area segmented as? (you can look at _class file generated by TOADS)

-Navid

Hi,

I think it is clear in the last view that the lower right part of gray matter is definitely lost. I also use FANTASM to segment, the result is attached, looking more reasonable.

Thanks,
Sophia

Hi,

I agree that it is clear from that image that gm is lost. I just wanted to see how the original image looks like in that area because it is not that clear from the sagital view. I have one recommendation and one request:

- Check the header of the image, make sure that the image orientation is correct. Both TOADS and SPECTRE require correct header (FANTASM does not). You can do this in mipav using Image->Attributes->view header (or Ctrl+H) . If it is wrong, you can correct it using Image->Attributes-> Edit attributes (or Ctrl+E). Make sure that you save the image after applying the corrections.

- Would you please attach the hard segmentation (*_seg file)generated by TOADS from that slice?

-Navid

Hi,

I think SPECTRE would reorient the image itself, right?
I corrected the image orientation and ran SPECTRE & TOADS again, still not working...

Thanks,
Sophia