Winston-Lutz errors

Hello Pylinac group,

I am trying to deliver a Winston Lutz test however I keep getting the following error codes:

  1. IndexError: index -3 is out of bounds for axis 0 with size 1
  2. ValueError: Unable to locate the BB. Make sure the field edges do not obscure the BB and that there is no artifacts in the images.

Details of the delivered beam and phantom are as followed

  • Varian Isocentre cube used (BB~2mm in diameter)
  • Beam delivered is 6MV 4x4cm.

Has anyone found success using the Varian Isocentre Cube as their phantom of choice? I imagine the image quality is causing Pylinac to struggle to find the BB but I am not sure how to rectify this.


Hi Hinesh,
See this discussion for the same problem: Winston Lutz module Error ( Pylinac by default only searches for BBs 3-10mm (I never used the Varian cube in the clinic). As stated there, this is solvable by changing the source code but will be easier in the future. If you’d like to make sure pylinac can handle such a scenario you can upload the images here and I’ll add them to the test suite:


Thanks for this, I tried changing the source code but it hasnt changed the errors. I have uploaded the files to your dropbox. Thanks again for your help :slight_smile:

Hi Pylinac group,

I am also trying to use the Varian isocenter cube phantom (BB~2mm in diameter), but I have the same problem “ValueError: Unable to locate the BB. Make sure the field edges…”.

Did you have success?

I am also using a 6MV beam, 10cm x 10cm field size, and a MV High Quality image acquisition template.

Pylinac 3.12.0

wl.analyze(bb_size_mm = 2)

As can be seen in the next two (arbitrary/example) images (pylinac demo on the left, my image on the right) I am not using a full open field, does that could be a problem?
I have loaded my dicom files to google from with this description “Varian Isocentre Cube. Winston-Lutz error (unable to locate the BB)”.



This is a hard problem to solve because of the very low contrast and small size of the BB. Even if I eliminate the size limitation of the BB, detecting the actual BB ROI is very difficult. E.g. here is the detected BB for one image and the resulting analysis. So, even if we loosen the algorithm the noise is an issue as stated in the release notes.



Hi James, thanks for your attention and fast answer.

What about if pylinac would have the option to allow users a manual identification? Something like a zoom cursor on the image? I know it is a lot of GUI-Canvas work (and maybe beyond pylinac), but if you think that could be a chance, I would like to work on that.



That’s really outside the scope of the project as pylinac is meant truly as a library only. Pylinac General Overview - pylinac 3.12.0 documentation
You are welcome to make a separate project that adds a GUI layer on top and if you open source it I’m sure the community would happily embrace it. E.g. this project used pylinac + orthanc. I’m happy to give input for it however.