Inconsistencies in ROI statistics outputs

Home › Forums › Protocols › Inconsistencies in ROI statistics outputs

Tagged: 

  •
    Madina Wane January 31, 2020 at 9:53 pm

    Hello,

    I have created a protocol with active contours which creates ROIs and export the areas to a workbook via the ROI statistics block. I have found that the area calculated by ROI statistics in the protocol is different to the area when running ROI statistics manually i.e. when I open the image/polygon ROIs from the ‘Display’ block in the protocol, and run ROI statistics from the toolbar.

    I am unsure why I am getting two different areas. It’s only a small difference (<1%) but I need to perform some additional manual operations with these ROIs so would like to understand why there is an inconsistency.

    Thank you!

    • This topic was modified 4 years, 10 months ago by
      Madina Wane      .
    • This topic was modified 4 years, 10 months ago by
      Madina Wane      .
    Stephane Dallongeville February 6, 2020 at 11:48 am

    Hi Madina,

    There is necessary something in between or maybe it’s a rounding issue.

    As you can see here, we obtain the same result from Protocol and the interactive plugin:

    So i guess there is something in the process (polygon to mask conversion) which may introduce smalls differences in your results. If you can share the protocol then we could eventually try to understand why this happen 🙂

    Best,

    – Stephane

     
    Madina Wane February 27, 2020 at 12:39 pm

    Hi Stephane,

    Apologies for the late response – I tried to sent a reply a few weeks ago but it looks like it never went through. I have tried several times to reply now but each time the message doesn’t appear on the thread. I’ve attached some screenshots of my protocol with an example of a processed image. I will try sending one screenshot at a time.

    Part 1:
    Madina Wane February 27, 2020 at 12:40 pm

    Here is part 2:
    Madina Wane February 27, 2020 at 12:42 pm

    Finally the output. I use the image generated in ‘Display 13’ to get the ROI stats manually. This is where I see the discrepancies.

    I am happy to send the full protocol file if that helps although I am not sure how to do this on the forum.

    Many thanks,

    Madina

    Stephane Dallongeville March 5, 2020 at 10:18 am

    Hi Madina,

    Thanks for your detailed reply and taking the time to show the protocol, unfortunately there is no way to attach protocol on the forum, it’s something we need to improve as it’s quite useful.

    But for now can you try to connect the binary image to your sequence input field from your ROI Statistics block as it requires it to correctly get the ‘pixel size’ metadata from the Sequence, maybe there is a small difference between the original image metadata and the binary image metadata that you manually set (rounding) which can explain the small difference between the 2.

    Best,

    – Stephane
    Madina Wane March 5, 2020 at 12:58 pm

    Hi Stephane,

    I’ve just tried this and it worked! The ROI statistics from the protocol and when manually obtained are identical. (In case you are wondeirng, the values match with the manually obtained ROI statistics from the screenshot above).

    Thank you very much for solving this issue!

    Best,

    Madina

    Stephane Dallongeville March 6, 2020 at 11:41 am

    Glad to hear that 🙂 You’re welcome !

    – Stephane

Viewing 8 posts - 1 through 8 (of 8 total)

The forum ‘Protocols’ is closed to new topics and replies.