Re: Question regarding the Virtual Insect Brain protocol

[dscho]

Hi Jitte,

best to discuss Open Source in the open; I Cc:ed the mailing list, please
use reply-to-all!

On Thu, 26 Jun 2014, Jitte Groothuis wrote:

> I am a PhD candidate at the Laboratory for Entomology at Wageningen
> University. My project involves brain morphology of *Nasonia *parasitoid
> wasps and I want to study effects of body size on brain scaling, and
> differences in the relative investment in different neuropils.  To do
> this I want to make "standard brains" for  different experimental groups
> and compare these.
>
> The Virtual Insect Brain protocol seems like a great tool for what I
> want to study, but unfortunately I have been running into some problems
> and was hoping that you might help me in clarifying some issues.

Unfortunately these are busy times for me because we are working hard on
finalizing ImageJ 2.0.0 (it's been quite an exciting time these past weeks
and months!).

Therefore I can spare only a couple of hours here and there to work on
this.

> I am primarily bothered by visualizing the output and the statistics
> involved in the VIB output. If you would be available I would be more
> than happy to send an additional e-mail with a more in-depth exploration
> of my issues.

If you could describe the problems you encountered and provide me with
(small) datasets to reproduce the problem, I am optimistic that I might be
able to help you!

Ciao,
Johannes

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

[Jitte Groothuis]

Dear Johannes and the ImageJ mailing list,

Thank you (all) for your taking the time to read this.
For clarity I will start off with most of the same information as I sent
Johannes in my original message:


I am a PhD candidate at the Laboratory for Entomology at Wageningen
University. My project involves brain morphology of *Nasonia* parasitoid
wasps and I want to study effects of body size on brain scaling, and
differences in the relative investment in different neuropils.
To do this I want to make "standard brains" for these different groups and
compare these.

The Virtual Insect Brain protocol seems like a great tool for what I want
to study, but unfortunately I have been running into some problems and was
hoping that you might help me in clarifying some issues.

Below I have outlined my questions, and have added a short overview of my
workflow. I will include a few examples of the label output that hopefully
illustrate my problem. As my questions feel more conceptual to me, and are
not so much about errors in VIB, I am not including a dataset.
Also included is a picture of a surface model based on an older
segmentation of a single brain. In a perfect world my standard brains would
look like this.


   1. Users of the VIB protocol (such as Kurylas et al, 2008) report e.g.
   volumes for their standard brains.
   As the VIB output does not seem to contain information or statistics
   about *average *neuropil volume or shape, and metadata such as the voxel
   size calibration is gone, I am at a loss how to acquire such data. Could it
   be because I have some problems opening the *.Statistics files?
   I have been assuming that these are not that important for this issue,
   as they pertain to individual brains and not the "standard brain".

   2. What is the proper way to visualize the standard brain from the VIB
   output?
   The output is one label file that contains probability data, and authors
   used this to recreate average fully segmented models. I understand this is
   mainly done by thresholding (one value in general, or different values for
   different neuropils) and subsequent re-segmentation.
   It seems to me that this would lead to new arbitrary choices and weird
   segmentation artifacts, such as jagged edges and empty space between
   neuropils, especially for neuropils enclosed in a larger volume (e.g. I
   include parts of the central complex, mushroom body, and lateral horn
   within the central brain)

   3. Tangential question: is the protocol (e.g. center of gravity
   calculations) influenced by small or large islands in the segmentation?
   As I mention above, I have for example a label for the larger part of
   the central brain, but there are large holes in it because of other
   neuropils (again: mushroom bodies, central complex, etc)



For reference: so far, my workflow is as follows (in a nutshell):

   1. *Prepare data* (dissection, nc82 staining, mounting in Vectashield,
   CLSM in several stacks);
   2. *Stitching* b-anterior and b-posterior stacks in FIJI (acquired this
   way due to limited working distance of the lens and too much signal
   attenuation even if it would fit);
   3. *Segmentation *of neuropils in AMIRA (because of interface
   preferences);
   4. *Converting *labelfiles to be similar to the output of the VIB
   stitcher (i.e. converting to tiff and saving as .Labels);
   5. *Run VIB* in FIJI (resampling 2, and choosing the nicest looking
   brain as template - for now I don't have a metric to calculate the best
   sample)


Thank you all again for your time and any help you can provide.
To all of you working on ImageJ 2.0.0: many thanks and keep up the good
work!

Sincerely,
Jitte

Jitte Groothuis, MSc

PhD Candidate

Wageningen UR

Laboratory for Entomology

P.O. Box 16, 6700 AA Wageningen

Wageningen Campus, building 107 (Radix)

Droevendaalsesteeg 1, 6708 PB Wageningen.

Tel. [phone] (desk)

Websites: LinkedIn <http://www.linkedin.com/pub/jitte-groothuis/41/36/376>
 / ResearchGate <http://www.researchgate.net/profile/Jitte_Groothuis/>


On 26 June 2014 17:27, Johannes Schindelin <[hidden email]>
wrote:

> Hi Jitte,
>
> best to discuss Open Source in the open; I Cc:ed the mailing list, please
> use reply-to-all!
>
> On Thu, 26 Jun 2014, Jitte Groothuis wrote:
>
> > I am a PhD candidate at the Laboratory for Entomology at Wageningen
> > University. My project involves brain morphology of *Nasonia *parasitoid
> > wasps and I want to study effects of body size on brain scaling, and
> > differences in the relative investment in different neuropils.  To do
> > this I want to make "standard brains" for  different experimental groups
> > and compare these.
> >
> > The Virtual Insect Brain protocol seems like a great tool for what I
> > want to study, but unfortunately I have been running into some problems
> > and was hoping that you might help me in clarifying some issues.
>
> Unfortunately these are busy times for me because we are working hard on
> finalizing ImageJ 2.0.0 (it's been quite an exciting time these past weeks
> and months!).
>
> Therefore I can spare only a couple of hours here and there to work on
> this.
>
> > I am primarily bothered by visualizing the output and the statistics
> > involved in the VIB output. If you would be available I would be more
> > than happy to send an additional e-mail with a more in-depth exploration
> > of my issues.
>
> If you could describe the problems you encountered and provide me with
> (small) datasets to reproduce the problem, I am optimistic that I might be
> able to help you!
>
> Ciao,
> Johannes
>
>

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

[Jitte Groothuis]

Dear ImageJ community,

The following message is a repost/resending of an earlier question
regarding the virtual insect brain protocol. I did not know at the time,
but, judging from most other listserv messages I read afterwards, I must
have sent my original message in quite a busy and difficult time for most
of you.



I am a PhD candidate at the Laboratory for Entomology at Wageningen
University. My project involves brain morphology of *Nasonia* parasitoid
wasps and I want to study effects of body size on brain scaling, and
differences in the relative investment in different neuropils.
To do this I want to make "standard brains" for these different groups and
compare these.

The Virtual Insect Brain protocol seems like a great tool for what I want
to study, but unfortunately I have been running into some problems and was
hoping that you might help me in clarifying some issues.

Below I have outlined my questions, and have added a short overview of my
workflow. I will include a few low-quality examples of the label output
that hopefully illustrate my problem. As my questions feel more conceptual
to me, and are not so much about errors in VIB, I am not including a
dataset.
Also included is a sample picture of a surface model based on an older
segmentation of a single brain. In a perfect world my standard brains would
look like this.


   1. Users of the VIB protocol (such as Kurylas et al, 2008) report e.g.
   volumes for their standard brains.
   As the VIB output does not seem to contain information or statistics
   about *average *neuropil volume or shape, and metadata such as the voxel
   size calibration is gone, I am at a loss how to acquire such data. Could it
   be because I have some problems opening the *.Statistics files?
   I have been assuming that these are not that important for this issue,
   as they pertain to individual brains and not the "standard brain".

   2. What is the proper way to visualize the standard brain from the VIB
   output?
   The output is one label file that contains probability data, and authors
   used this to recreate average fully segmented models. I understand this is
   mainly done by thresholding (one value in general for all neuropils, or
   different values for different neuropils) and subsequent re-segmentation.
   It seems to me that this would lead to new arbitrary choices and weird
   segmentation artifacts, such as jagged edges and empty space between
   neuropils, especially for neuropils enclosed in a larger volume (e.g. I
   include parts of the central complex, mushroom body, and lateral horn
   within the central brain)

   3. Tangential question: is the protocol (e.g. center of gravity
   calculations) influenced by small or large islands in the segmentation?
   As I mention above, I have for example a label for the larger part of
   the central brain, but there are large holes in it because of other
   neuropils (again: mushroom bodies, central complex, etc)



For reference: so far, my workflow is as follows (in a nutshell):

   1. *Prepare data* (dissection, nc82 staining, mounting in Vectashield,
   CLSM in several stacks);
   2. *Stitching* b-anterior and b-posterior stacks in FIJI (acquired this
   way due to limited working distance of the lens and too much signal
   attenuation even if it would fit);
   3. *Segmentation *of neuropils in AMIRA (because of interface
   preferences);
   4. *Converting *labelfiles to be similar to the output of the VIB
   stitcher (i.e. converting to tiff and saving as .Labels);
   5. *Run VIB* in FIJI (resampling 2, and choosing the nicest looking
   brain as template - for now I don't have a metric to calculate the best
   sample)


Thank you all again for your time and any help you can provide.
To all of you working on ImageJ/FIJI: many thanks and keep up the good work!

Sincerely,
Jitte

Jitte Groothuis, MSc

PhD Candidate

Wageningen UR

Laboratory for Entomology

P.O. Box 16, 6700 AA Wageningen

Wageningen Campus, building 107 (Radix)

Droevendaalsesteeg 1, 6708 PB Wageningen.

Tel. [phone] (desk)

Websites: LinkedIn <http://www.linkedin.com/pub/jitte-groothuis/41/36/376>
 / ResearchGate <http://www.researchgate.net/profile/Jitte_Groothuis/>


On 26 June 2014 17:27, Johannes Schindelin <[hidden email]>
wrote:

> Hi Jitte,
>
> best to discuss Open Source in the open; I Cc:ed the mailing list, please
> use reply-to-all!
>
> On Thu, 26 Jun 2014, Jitte Groothuis wrote:
>
> > I am a PhD candidate at the Laboratory for Entomology at Wageningen
> > University. My project involves brain morphology of *Nasonia *parasitoid
> > wasps and I want to study effects of body size on brain scaling, and
> > differences in the relative investment in different neuropils.  To do
> > this I want to make "standard brains" for  different experimental groups
> > and compare these.
> >
> > The Virtual Insect Brain protocol seems like a great tool for what I
> > want to study, but unfortunately I have been running into some problems
> > and was hoping that you might help me in clarifying some issues.
>
> Unfortunately these are busy times for me because we are working hard on
> finalizing ImageJ 2.0.0 (it's been quite an exciting time these past weeks
> and months!).
>
> Therefore I can spare only a couple of hours here and there to work on
> this.
>
> > I am primarily bothered by visualizing the output and the statistics
> > involved in the VIB output. If you would be available I would be more
> > than happy to send an additional e-mail with a more in-depth exploration
> > of my issues.
>
> If you could describe the problems you encountered and provide me with
> (small) datasets to reproduce the problem, I am optimistic that I might be
> able to help you!
>
> Ciao,
> Johannes
>
>

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