[ITK-users] [ITK] adding a global hierarchical watershed segmentation mode to itksnap

Tue Aug 12 08:42:31 EDT 2014

Hi Roman,

Try GrowCut segmentation. You paint a few strokes inside and outside your target(s) on one or a few slices and you may get a nice sementation. Using the FastGrowCut version you can edit/refine your inputs and as it is GPU-accelerated, it's fast (initial segmentation may take 20-30 seconds on a high-resolution volume but then the updates take just a few seconds).

I've just used ITKSnap a couple of times, so I don't know it much. Slicer is a large, more comprehensive application framework, with a large user and developer community. You can add new segmentation methods to Slicer by writing a Python script that can use VTK, ITK, numpy, etc. No developer tools are needed, no need to compile anything, just download Slicer and you are ready to go. Of course you can also develop in C++, if you prefer; and if you download Eclipse then you can do step-by-step debugging of your Python module in Slicer, etc. You can share your module with others by uploading it to the extension manager (Slicer app store). More info: http://www.slicer.org/slicerWiki/index.php/Documentation/Nightly/Developers.

Andras

-----Original Message-----
From: Dr. Roman Grothausmann [mailto:grothausmann.roman at mh-hannover.de] 
Sent: Tuesday, August 12, 2014 4:15 AM
To: Andras Lasso
Cc: itksnap-users at googlegroups.com; ITK Mailing List
Subject: Re: [ITK] [ITK-users] adding a global hierarchical watershed segmentation mode to itksnap

Hi Andras,

Many thanks for Your hints. We'll test them on our data. However, our data is not as standardized in quality as CT or MRT data and from our experience many of the "magic" segmentations demonstrated on CT or MRT data does not work on our data.
However, of all the segmentation modules I could find in Slicer or elsewhere none was based on a hierarchical watershed algorithm like the SegmentationEditor of the InsightApplications. Is there any other that I've missed?
Concerning my idea to implement this if no tool currently already exists:
Would such an extension be of more use in Slicer or itksnap?
For which of the both would it likely be easier to integrate the extension?

Best wishes
Roman

On 11/08/14 16:20, Andras Lasso wrote:
> Have you tried segmentation tools in 3D Slicer? There is a multi-label 
> watershed segmentation tool that works in 3d 
> (http://www.slicer.org/slicerWiki/index.php/Documentation/Nightly/Modu
> les/Editor#WatershedFromMarkers), there are also a number of other 
> generic semi-automatic segmentation tools (such as GrowCut) and a few 
> fully-automatic specialized tools (for brain, airways, etc). Download 
> the nightly build and also checkout the Segmentation category in the 
> Extension Manager.
>
> Andras
>
> *From:* Dr. Roman Grothausmann 
> <mailto:grothausmann.roman at mh-hannover.de>
> *Sent:* ‎Monday‎, ‎August‎ ‎11‎, ‎2014 ‎1‎:‎15‎ ‎AM
> *To:* itksnap-users at googlegroups.com 
> <mailto:itksnap-users at googlegroups.com>
> *Cc:* ITK Mailing List <mailto:insight-users at itk.org>
>
> Dear mailing list members,
>
>
> For a long time now I've been searching for a replacement of the old 
> SegmentationEditor (part of the InsightApplications) which I could not 
> get to work even with older ITK-3.x. Somewhere I was told that it was 
> integrated and maintained within itksnap. However the only tool in 
> itksnap based on watershed segmentation seems to be the adaptive 
> brush. It is very handy, but time-consuming in 3D and only for local 
> segmentations. In contrast the SegmentationEditor used a global 
> hierarchical watershed segmentation to pick labels from, see e.g. the video linked here:
> http://www.insight-journal.org/browse/publication/70
> Quite often now I've stumbled over data, that would likely have 
> profited a lot of such a tool for segmentation.
>
> If such a tool working with ITK-4.x does not exist any more, I thought 
> of implementing it as a global WS segmentation mode in itksnap. That 
> way making the tool easier to find, to use and to ensure maintenance 
> (should it be integrated into the itksnap code base in the end) and 
> also avoiding recoding a GUI with mouse interaction and such.
>
> My current idea is to start, similar to the snake-tool, with 
> pre-filtering like smoothing and optional edge-detection.
> Then the main edit mode to consist of the usual 4 subwindows showing 
> an overlay of the grey-image and the watershed segmentation of the 
> hierarchical WS-filter (hWS), side-by-side to another label-dataset 
> (loaded like a tiled overlay) which consist of the actually chosen labels from the hWS as the final data (FD).
> The interaction I'm considering would be to LMB click on a label in 
> the hWS which is then added to the FD as the current fg label, or removed with a RMB.
> The hierachrchy level of the hWS could be controlled by a slider or 
> the mouse wheel + Shift in nearly real time if some region demands 
> higher or lower detailed segmentation. When the fg label is changed, 
> the segmentation added or removed with further mouse interactions would modify the new fg label.
> When all is done and the global WS mode is left, the FD label-dataset 
> is added to the main segmentation of itksnap.
>
> Would that be possible to implement in itksnap 3.x?
> What source files of the itksnap code base should I look at for:
> - the pre-filtering mode of the snake (for the smoothing and edge 
> detection)
> - the mouse interactions (for e.g. adding the whell + Shift 
> interaction)
> - possible examples for the dual layout
>
> Thanks for any comments, help or hints.
> Roman
>
> --
> Dr. Roman Grothausmann
>
> Tomographie und Digitale Bildverarbeitung Tomography and Digital Image 
> Analysis
>
> Institut für Funktionelle und Angewandte Anatomie, OE 4120 
> Medizinische Hochschule Hannover Carl-Neuberg-Str. 1
> D-30625 Hannover
>
> Tel. +49 511 532-9574
> _____________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at 
> http://www.kitware.com/opensource/opensource.html
>
> Kitware offers ITK Training Courses, for more information visit:
> http://www.kitware.com/products/protraining.php
>
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/insight-users
> _______________________________________________
> Community mailing list
> Community at itk.org
> http://public.kitware.com/mailman/listinfo/community

--
Dr. Roman Grothausmann

Tomographie und Digitale Bildverarbeitung Tomography and Digital Image Analysis

Institut für Funktionelle und Angewandte Anatomie, OE 4120 Medizinische Hochschule Hannover Carl-Neuberg-Str. 1
D-30625 Hannover

Tel. +49 511 532-9574