- Queries
- All Stories
- Search
- Advanced Search
- Transactions
- Transaction Logs
All Stories
May 10 2021
May 8 2021
After talking with Marco, I have looked for the whereabouts of dockerfiles (liver-mri and multiorgan-seg), working commits for MITK (a0dedffb74d) and for MBI-builds (56ad5bf7064) and trained models. @schererj: I sent you this information via email in October 2020, and shared the trained models on the network drives. Can you confirm? What information do you need to keep the shape models operational?
May 7 2021
In T28332#223158, @kleina wrote:I updated the draft in the branch directly (same file).
- I did not change the send/recv to scatter/gather. In my opinion send/recv makes more sense, but we can discuss this.
- In my opinion, we maybe don't need the "cleaning up" up the mpi_processes here, because it might be taken care of by our current implementation.
Let me know when you had time to look at it and if there are any questions. I'd like to discuss it a bit.
Could you point me to the commit? I cannot find it ?!?
Hm, due to the facts that (1) the feature is missing since 2016 and no one ever complaint, (2) it is not documented in the help and (3) I am not sure why one would like to behave differently when clicking in an area that is segmented, I would just skip the code path. Yes.
But I would go further then. You don't need the whole inside check any more. I would completly remove everythin below line 332 and directly call OnMousePressedOutside(nullptr, interactionEvent); then.
I've started a draft for a message to send around, found here if you want to have a look or add something: https://hub.dkfz.de/s/HzpkipeWRkq3K8K
I quickly modified the code to see if this works potentially. For this I set m_PaintingPixelValue to 1 for both cases and used the OnMousePressedOutside-function also for a mouse pressed inside:
m_PaintingPixelValue = inside ? 1 : 1;
I started looking into this and what I found inside mitk::RegionGrowingTool::OnMousePressed:
if (inside) { MITK_DEBUG << "Clicked inside segmentation"; // For now, we're doing nothing when the user clicks inside the segmentation. Behaviour can be implemented via // OnMousePressedInside() // When you do, be sure to remove the m_PaintingPixelValue check in OnMouseMoved() and OnMouseReleased() return; } else { MITK_DEBUG << "Clicked outside of segmentation"; OnMousePressedOutside(nullptr, interactionEvent); }
Looking at the file history it doesn't look like that was implemented in MITK at anytime.
So we can actually discuss if we want to allow such a functionality and if, what the obstacles are.
May 6 2021
Obviously it was not clear what we are expecting here so I will close this task and refer to two new tasks for the specific topics mentioned here:
- Where to put "publicly available" / "shareable" test data --> availability: T28480: [Checklist] [Test data] Define publicly available location to store user test data
- Which data to use for specific checklists --> suitability: T28481: [Checklist] [Test data] Define publicly available user test data
May 5 2021
Maybe something to keep an eye on: T28478 and https://github.com/InsightSoftwareConsortium/ITK/tree/master/Modules/ThirdParty/DCMTK
May 4 2021
Thanks for the clarification. I agree 100%.
Don't get me wrong - I'm not saying that you or someone "verschlimmbesserte" anything :D I understand and appreciate all the effort that has been made by you.
I am more thinking of a different approach architecture wise - you already mentioned "abstraction levels". For me it seems as if different classes and functions are on the same "level" of doing something but they do it differently - which makes it even harder to grasp the logic behind it. That's why I mentioned "class diagrams". Also, often classes and functions are way too overloaded (e.g. SegTool2D) with way too many responsibilities.
In T28131#223675, @kalali wrote:Ok I have to admit: The whole code base that is related to this topic is totally unclear to me and I find it very hard to understand what is going on.[...]
Told you so. ;) As proposed we can have a session together.
Ok I have to admit: The whole code base that is related to this topic is totally unclear to me and I find it very hard to understand what is going on. I don't know why the code is like that but I would highly suggest to simplify and clean the relevant classes and functions as much as possible before it becomes problematic (e.g. more complex cases with 4D-dynamic segmentations with multi-layer multi-label segmentation masks or similar). Also, I highly doubt that this is future proof when it comes to changes in the following years so as already mentioned in my last comment: I think we really need to discuss this topic - maybe it helps to have a clearer picture by creating (class) diagrams with relations etc.
Thanks for the clarification. I will look into the FeedbackCntourTool and ContourModelUtils and see if I can make sense of all this :D
Depending on what I find out I think we should really talk about the future of the (multilabel) segmentation modules / plugins and how we can simplify the code base (generalize it) and remove redundancy as well as decouple classes etc.
May 3 2021
In T28206#223651, @kalali wrote:
You meen T28221, don't you?
Great you are looking into that matter.
The problem is caused by mitk::AutoSegmentationWithPreviewTool::TransferImageAtTimeStep() more precisely in the private helper template function ITKSetVolume (same cpp).
I am not sure what is exactly done here to reproduce. To get any help page I have to click on a plugin anyway. How would I provoke a blank page?
When I use F1 while the Standard Display is activated I get a blank page for the Standard Display. If that's what is meant than we need to reconsider how to display this specific help page: There is a description of the "four window view" inside src\Plugins\org.mitk.gui.qt.ext\documentation\UserManual\MITKUserManual.dox but we can not move it to the stdmultiwidgeteditor-plugin as it contains more information about the workbench, not only about the standard display.
I tried to reproduce this issue and found the following:
With the changes in T28211 @floca mentioned, there is no legend anymore m_Controls.chartWidget->SetShowLegend(false);
Showing this legend again displays "histogram" as the label name - not even the original segmentation node name. Now when I add another segmentation and add it to the statistics, the histogram is completely removed (it's calculated but the widget does not provide any space for the histogram anymore).
@floca I sarted looking into this and found out the following:
The threshold tool used is QmitkBinaryThresholdToolGUIBase, which is a subclass of QmitkAutoSegmentationToolGUIBase. This class uses the AutoSegmentationTool, a subclass of the mitkTool. As far as I can see the target node that will be connected with this tool upon activating the tool is done inside mitk::AutoSegmentationTool::GetTargetSegmentationNode().
However, there is no information about any layer or even the information that this can be used with multilabel segmentation nodes.
Let me know if you got hold of it.
Apr 30 2021
Deleted branch from rMITK MITK: release/T28473-2021-Week-17.
Pushed new branch to rMITK MITK: release/T28473-2021-Week-17.
Apr 29 2021
Apr 28 2021
Deleted branch from rMITK MITK: feature/T28118-Refactor_SegTool2D_and_derived_classes.
@floca Please delete the feature branch mentioned above if this task is resolved.
Deleted branch from rMITK MITK: bugfix/T28437-RegionGrowing3DSeedPoint0.