[leonidastra86]

Is there a way to accelerate the procedure? I need sometimes 12-14 hours for just one patient.
Do I need more Ram or a better processor?

Is there a way to calculate also tumor-volumes? Or is it made just for normal brain? The results of the segmentation by patients with tumors, are they accurate?

[Anand Joshi]

It should take 2-3 hrs for BrainSuite+SvReg on normal brains if there is sufficient memory and processing power available. Please make sure that there is >=4GB RAM free when you are running the software. For this purpose, closing as many programs as possible is helpful. Also make sure that hibernation or sleep modes are disabled.

[leonidastra86]

I have already tried it in my Macbook with Sierra and my PC with Windows 10, the first one has 8 GB Ram and the second 16. In both cases I have needed many hours for it. Is there something I should change in the configuration of the program? Does it have to do maybe with the patients, which all had a tumor? Does it work only with high resolution MRIs?
Thank you for your time and responses!

[Anand Joshi]

My guess is that due to tumor the surfaces generated by BrainSuite have too many faces or vertices. This is causing slowing down of the processes. You can try clicking surface registration only checkbox at the beginning of svreg sequence if you need only surface labels.
What is the voxel resolution of the MRIs? As long as it is around 1mm it should not matter too much.

• This reply was modified 7 years, 6 months ago by Anand Joshi. Reason: edits

[leonidastra86]

1mm, its not so much.

[Anand Joshi]

ok. it is probably related to the surface generated by BrainSuite is too complicated due to tumor. It would depend on shape of the tumor. I suggest first trying a normal brain. One example can be found here:
http://brainsuite.org/tutorials/cseexcercise/
If this brain also takes a lot of time then there is something wrong with the setup otherwise it is related to the tumor related cortical extraction issues.

[leonidastra86]

So, brain suite cannot be used for brain tumors. Is that right?
What is the range of the voxel resolution which works with brain suite?

[Anand Joshi]

There is no limitation on voxel size but 1mm is generally optimal for T1 brain scans. In our experience, BrainSuite generally works with tumor brains in the sense that the labeling of regions away from the tumor region is relatively unaffected. The execution time can vary for the tumor cases though it should not be too high. Did you get a chance to try normal brains with BrainSuite to make sure that the long execution time on your computer is not related to tumors?
If possible, it would also be helpful if you can post a screenshot of your results with the tumor.

[leonidastra86]

With a normal brain I was going to try it today.
I have already done it with many brains with tumor and the results are the following
http://i63.tinypic.com/o8f9m9.jpg
http://i66.tinypic.com/dxmzcg.jpg
http://i66.tinypic.com/9uvm6c.jpg
http://i65.tinypic.com/15ev6gl.jpg
http://i65.tinypic.com/1z1tsgy.jpg
As we see, the tumor is missing region. I cannot (directly at least) calculate its size or position.
Furthermore, the tumor is shown normally in the 3D projection and not on the other side (like in mirror) which is mostly in MRI the case.
Last but not least, it would be really interesting to have at the and of the analysis, at least the localisation of the tumor (parietal, central region, or much better more accurately specified). Thats shouldn’t be a problem for the software, as it is enough to compare the normal side and its structures with the tumor side.

[leonidastra86]

With the normal brain it took only 2 1/2 hours as you said.

[Author]

Posts