Header

Segmentation Fault: 11

Submitted by willi3by on
Category

Hi,

I am trying to run DIFFPREP and I am getting the following error message:

Denoising DWIs...

Segmentation Fault: 11

 

The data that I am using is 220 total volumes with 3 shells (1000, 2000, 3000). I know this error is sometimes caused by memory issues but I have 32GB of RAM, which should be plenty I think. Does anyone know what may be causing the issue? 

 

Thanks,

Brady

Submitted by willi3by on Thu, 09/19/2019 - 16:23

Permalink

I wanted to mention that this does not occur 100% of the time. If I denoise before starting DIFFPREP and set -d to off, then I only get this error about 95% of the time after the "Estimating noise levels.." step. Any guidance is greatly appreciated. 

 

Thanks

Submitted by irfanogl on Thu, 09/19/2019 - 16:29

Permalink

Okay.. 

I think you are using a version that is older than 3.1.4.   Before that version, the matrix PCA was performed using the vnl library, which sometimes gave this weird seg faults.  beginning with 3.1.4, all the matrix operations were changed to Eigen which is much more stable.

 

So can you confirm your version? If it is older, can you try with the newest version and let us know?

 

Thank you for the response. I am using the latest version, TORTOISE 3.1.4. I just installed on the computer that I currently using in the middle of July of this year. 

Thanks,

Brady

Submitted by willi3by on Thu, 09/19/2019 - 16:33

Permalink

Thank you for the response. I am using the latest version, TORTOISE 3.1.4. I just installed on the computer that I currently using in the middle of July of this year. 

Thanks,

Brady

Submitted by irfanogl on Thu, 09/19/2019 - 17:14

Permalink

Okay that is pretty weird then. I have never seen that seg fault happen with the Eigen based version.

If you guys are okay anonymizing the data and sharing the NIFTI file with us, we can take a look.

Submitted by willi3by on Sat, 09/21/2019 - 09:46

Permalink

Sure, what is the best way to share the files with you? The data will be about 1 GB total. I have been using TORTOISE for about 4 years and this is the first time this has happened. The only difference between this dataset and previous datasets is the size. This dataset is 3 shells with 220 total volumes whereas our previous data has been 1 shell with a maximum of 61 directions. Also, this is the first time I've tried to run it on MacOS 10.14 Mojave. Not sure if that has something to do with it. Thanks for looking into it.

 

Brady

Submitted by willi3by on Sat, 09/21/2019 - 14:47

Permalink

Also, I just thought about the fact that DWIDenoise runs successfully every time. When using DIFFPREP, is a different denoising algorithm used to estimate noise levels? If not, I'm not sure why it would fail with DIFFPREP but work with DWIDenoise. 

 

Thanks again,

Brady

Submitted by irfanogl on Mon, 09/23/2019 - 11:20

Permalink

No they are 100% identical codes. That is why it is weird and I have to see the data.

 

If you can send me an email at irfanoglumo@nih.gov   , I will send you a link where you can put the data.