removing false-positives

[asud1]

I've been using the Mac version of AxonSeg, and it all works really well until the last part (Step 3):

I don't seem to be able to remove most any of the false-positives, and when I click "Use Discriminant Analysis", it seems to crash the program (it processes for over 30 mins without completing; I have 32GB ram, so memory shouldn't be a problem).

Also, the function to add-in false-negatives doesn't seem to be there (I was following this video tutorial, and it seems to have been a feature previously: https://www.youtube.com/watch?v=c-Vi5MHYFIM ).

Is there a fix for these issues?

Thanks for the help!

[alzaia]

Hello,

Issue 1 ---

Does the GUI crash if you do not use the discriminant analysis option (i.e. simply segment myelin without doing any discriminant analysis)?

My suggestion would be to first use the software without the discriminant analysis option, and check the results you get.

Issue 2 ---

About the possibility to add false negatives, we decided to remove than option when we published, and explained our decision in the paper as follows:

Note that we do not let the user add missing axons in the GUI at this step for two main reasons. First, manually adding missed axons can add bias in shapes and thus strongly affect the DA. Secondly, AxonSeg comes with a smaller GUI tool called ManualCorrectionGUI, which can be used in order to correct a segmentation result by adding, removing or modifying axons after the full-image segmentation.

There is another solution for that. You can actually correct your axonal mask (i.e. add missing axons or remove false positives) on your favourite image processing software (ImageJ, GIMP, ...) and feed it back again into the AxonSeg pipeline to get the myelin segmentation based on the corrected axon mask. The procedure is explained in the AxonSeg tutorial script (PART 5).

Note:

Depending on what your application is, you may want to consider trying our latest software AxonDeepSeg. It uses deep learning to automatically segment the axons and myelin.

Cheers!

[asud1]

Hi, thanks for answering me so quickly!

Issue 1 - no it doesn't crash when I don't run the discriminant analysis. It completes the segmentation fine.

Thanks for forwarding the tutorial for feeding amended images back in, and the link to the new software. I hadn't come across these, and will try both

Thanks for all your help!