Performing manual segmentation

[Kevin Keraudren]

Hi,

I would like to manually segment a few cell images in order to evaluate
an automated segmentation (for instance trace the cell boundaries and
the contour of the nuclei).

I usually do manual segmentations with ITK-snap on medical images, not
microscopy images, but I think it is overkill for PNG/TIFF images, and
not appropriate to handle classes: I would like to have the cytoplasm
and the nuclei classes, and assign labels of the form cytoplasm-1,
cytoplasm-2, cytoplasm-3,... and nuclei-1, nuclei-2, nuclei-3.
For my application, I would have more classes than just cytoplasm and
nuclei.

Could I use ImageJ for this? Otherwise would you know of an appropriate
program for this kind of manual segmentation. In my mind it would be a
tool that allow you to segment the image with layers, as the layers of
Photoshop or GIMP, and each layer would correspond to a class.

Thanks a lot for your help,

Kind regards,

Kevin

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Jerome Mutterer-3]

Hi Kevin,

You could take advantage of the recent possibility to assign properties to
ROIs in ImageJ 1.48h.

I put together a simple test macro that does this. You can get it at:
https://gist.github.com/mutterer/7758228#file-roiclassestest-ijm
and adapt it to your needs. It just assigns a custom property and a stroke
color to the active ROI, and pushes it to the ROI manager. Then you can
select ROIs by keyword, which could be a starting point for your manual
segmentation analysis.

I recorded a small video about how to use it at:
http://youtu.be/RlhG23xPEbg

Sincerely,
Jerome



On 30 November 2013 14:22, Kevin Keraudren <[hidden email]
> wrote:

> Hi,
>
> I would like to manually segment a few cell images in order to evaluate an
> automated segmentation (for instance trace the cell boundaries and the
> contour of the nuclei).
>
> I usually do manual segmentations with ITK-snap on medical images, not
> microscopy images, but I think it is overkill for PNG/TIFF images, and not
> appropriate to handle classes: I would like to have the cytoplasm and the
> nuclei classes, and assign labels of the form cytoplasm-1, cytoplasm-2,
> cytoplasm-3,... and nuclei-1, nuclei-2, nuclei-3.
> For my application, I would have more classes than just cytoplasm and
> nuclei.
>
> Could I use ImageJ for this? Otherwise would you know of an appropriate
> program for this kind of manual segmentation. In my mind it would be a tool
> that allow you to segment the image with layers, as the layers of Photoshop
> or GIMP, and each layer would correspond to a class.
>
> Thanks a lot for your help,
>
> Kind regards,
>
> Kevin
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[lechristophe]

Woah! Thanks for sharing this.

ROI properties and tags are an excellent addition. I used to append text to
ROI names in the RoiManager to do that, but it a lot nicer and simpler to
have it integrated in ImageJ. Thanks Jerome!

Christophe

--
Christophe Leterrier
Chercheur
Equipe Architecture des Domaines Axonaux
CRN2M CNRS UMR 7286 - Aix Marseille Université



On Mon, Dec 2, 2013 at 10:12 PM, Jerome Mutterer <
[hidden email]> wrote:

> Hi Kevin,
>
> You could take advantage of the recent possibility to assign properties to
> ROIs in ImageJ 1.48h.
>
> I put together a simple test macro that does this. You can get it at:
> https://gist.github.com/mutterer/7758228#file-roiclassestest-ijm
> and adapt it to your needs. It just assigns a custom property and a stroke
> color to the active ROI, and pushes it to the ROI manager. Then you can
> select ROIs by keyword, which could be a starting point for your manual
> segmentation analysis.
>
> I recorded a small video about how to use it at:
> http://youtu.be/RlhG23xPEbg
>
> Sincerely,
> Jerome
>
>
>
> On 30 November 2013 14:22, Kevin Keraudren <
> [hidden email]
> > wrote:
>
> > Hi,
> >
> > I would like to manually segment a few cell images in order to evaluate
> an
> > automated segmentation (for instance trace the cell boundaries and the
> > contour of the nuclei).
> >
> > I usually do manual segmentations with ITK-snap on medical images, not
> > microscopy images, but I think it is overkill for PNG/TIFF images, and
> not
> > appropriate to handle classes: I would like to have the cytoplasm and the
> > nuclei classes, and assign labels of the form cytoplasm-1, cytoplasm-2,
> > cytoplasm-3,... and nuclei-1, nuclei-2, nuclei-3.
> > For my application, I would have more classes than just cytoplasm and
> > nuclei.
> >
> > Could I use ImageJ for this? Otherwise would you know of an appropriate
> > program for this kind of manual segmentation. In my mind it would be a
> tool
> > that allow you to segment the image with layers, as the layers of
> Photoshop
> > or GIMP, and each layer would correspond to a class.
> >
> > Thanks a lot for your help,
> >
> > Kind regards,
> >
> > Kevin
> >
> > --
> > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> >
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Kevin Keraudren]

Hi Jerome,

Thanks a lot for the gist, I'll give it a try.

The video does not seem to be working, Youtube says "This video is
currently unavailable".

On another mailing list (ITK), I've been advised to use TrackEM2 for
manual segmentation with ImageJ/Fiji.

I'll look into it as well.

Kind regards,

Kevin



Le 02/12/2013 21:12, Jerome Mutterer a écrit :

> Hi Kevin,
>
> You could take advantage of the recent possibility to assign properties to
> ROIs in ImageJ 1.48h.
>
> I put together a simple test macro that does this. You can get it at:
> https://gist.github.com/mutterer/7758228#file-roiclassestest-ijm
> and adapt it to your needs. It just assigns a custom property and a stroke
> color to the active ROI, and pushes it to the ROI manager. Then you can
> select ROIs by keyword, which could be a starting point for your manual
> segmentation analysis.
>
> I recorded a small video about how to use it at:
> http://youtu.be/RlhG23xPEbg
>
> Sincerely,
> Jerome
>
>
>
> On 30 November 2013 14:22, Kevin Keraudren <[hidden email]
>> wrote:
>> Hi,
>>
>> I would like to manually segment a few cell images in order to evaluate an
>> automated segmentation (for instance trace the cell boundaries and the
>> contour of the nuclei).
>>
>> I usually do manual segmentations with ITK-snap on medical images, not
>> microscopy images, but I think it is overkill for PNG/TIFF images, and not
>> appropriate to handle classes: I would like to have the cytoplasm and the
>> nuclei classes, and assign labels of the form cytoplasm-1, cytoplasm-2,
>> cytoplasm-3,... and nuclei-1, nuclei-2, nuclei-3.
>> For my application, I would have more classes than just cytoplasm and
>> nuclei.
>>
>> Could I use ImageJ for this? Otherwise would you know of an appropriate
>> program for this kind of manual segmentation. In my mind it would be a tool
>> that allow you to segment the image with layers, as the layers of Photoshop
>> or GIMP, and each layer would correspond to a class.
>>
>> Thanks a lot for your help,
>>
>> Kind regards,
>>
>> Kevin
>>
>> --
>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html