Questions regarding Sholl Analysis output

[jstanko]

FYI I'm using PC format, not Mac.

What is the difference between X0 Y0 and X1 Y1 values?

How can the analysis return a negative number of intersections?  

I'm using the plugin to count intersections in skeletonized images of rodent mammary glands and these are considerably more complex than a typical neuronal arbor.  Status indicates I'm reaching the max number of iterations resulting in (inaccurate result?) and I'm not really sure what to do here.  I would also like to utilize the radius span to minimize tangential intersection counts, but I'm not clear on how this works so I don't know what values to test.  Any help on these points would be appreciated.

Is it possible to save the graph of intersections/radii?  When I try to save it, it saves only the values, not the graph.

Thanks!

[Tiago Ferreira-2]

Dear Jason,

I will assume you are referring to the last version (2.1) of the Advanced Sholl
Analysis plugin <http://imagejdocu.tudor.lu/doku.php?id=plugin:analysis:asa:start>
and not the Sholl function of Fiji's Simple Neurite Tracer.

> What is the difference between X0 Y0 and X1 Y1 values?
X0 Y0 list the actual intersection counts.
X1 Y1 are the coordinates of the fitted curve.

> How can the analysis return a negative number of intersections?
It can't. But if you fitted the linear Sholl to a polynomial, that polynomial
can have negative values between starting radius and ending radius.

> I'm using the plugin to count intersections in skeletonized images of rodent
> mammary glands and these are considerably more complex than a typical
> neuronal arbor.  Status indicates I'm reaching the max number of iterations
> resulting in (inaccurate result?) and I'm not really sure what to do here.
These messages refer to the curve fitting. If none of the default fitting options
is suitable for your arbors and you still would like to perform the fit, uncheck
"Fit profile and compute descriptors", then copy the profile into 'Analyze>Tools>
Curve Fitting...' and perform a custom fit with IJ's built-in CurveFitter.
The CurveFitter has been recently revamped and while the documentation for the
new version is underway, you can have a look at the guide for IJ 1.45
<http://imagej.nih.gov/ij/docs/guide/userguide-27.htm>

But if you are already dealing with skeletons, did you consider using Ignatio's
Analyze Skeleton (in Fiji, Plugins>Skeleton>Analyze Skeleton),
<http://fiji.sc/wiki/index.php/AnalyzeSkeleton>?

> I would also like to utilize the radius span to minimize tangential
> intersection counts, but I'm not clear on how this works
See the documentation at
<http://imagejdocu.tudor.lu/doku.php?id=plugin:analysis:asa:start>
See also Tom's original documentation at
<http://www-biology.ucsd.edu/labs/ghosh/software/ShollAnalysis.pdf>

> Is it possible to save the graph of intersections/radii?  When I try to save
> it, it saves only the values, not the graph.
Use File>Save As>PNG... (or any other image format of your choice) to save the
canvas of any IJ plot.

If you think it helps, feel free to point us to a representative image

HTH,
-tiago

[jstanko]

Tiago,

 

Thanks for your response.  I’ve attached a rat mammary gland whole mount, skeletonized image, Sholl mask, and output files for one of my samples.  I’m trying to develop a quantitative method for describing the branching density of the gland and this method seems to be working well, just have to get the kinks worked out.  If you have any suggestions/recommendations, I’m open to anything that would help.

 

I thought the negative values may have been part of the polynomial, but since I wasn’t sure of the diff b/n XY0 and XY1, I wasn’t sure.  Makes sense.

 

The R2 values are in the 0.96-0.98 range for the images I’ve run so far so I’m not sure I need to do any additional curve fitting, though I am using 7^th order polynomial.  Unless you feel otherwise.

 

What is the advantage of Ignatio’s vs. Sholl?

 

I’ve read the Radius Span descriptions, I guess I’m not clear on how measurements are taken.  Let’s say for example a branch of the mammary gland runs tangent to the circumference of a circle for 100um (125 pixels/mm scale) , I is this counted as one intersection?  Or would I need to set the radius span to 0 or something else?  Or am I confused?

 

Thanks,

Jason

 

From: Tiago Ferreira-2 [via ImageJ] [mailto:[hidden email]]
Sent: Wednesday, May 02, 2012 12:47 PM
To: Stanko, Jason (NIH/NIEHS) [E]
Subject: Re: Questions regarding Sholl Analysis output

 

Dear Jason,

I will assume you are referring to the last version (2.1) of the Advanced Sholl
Analysis plugin <http://imagejdocu.tudor.lu/doku.php?id=plugin:analysis:asa:start>
and not the Sholl function of Fiji's Simple Neurite Tracer.

> What is the difference between X0 Y0 and X1 Y1 values?
X0 Y0 list the actual intersection counts.
X1 Y1 are the coordinates of the fitted curve.

> How can the analysis return a negative number of intersections?
It can't. But if you fitted the linear Sholl to a polynomial, that polynomial
can have negative values between starting radius and ending radius.

> I'm using the plugin to count intersections in skeletonized images of rodent
> mammary glands and these are considerably more complex than a typical
> neuronal arbor.  Status indicates I'm reaching the max number of iterations
> resulting in (inaccurate result?) and I'm not really sure what to do here.
These messages refer to the curve fitting. If none of the default fitting options
is suitable for your arbors and you still would like to perform the fit, uncheck
"Fit profile and compute descriptors", then copy the profile into 'Analyze>Tools>
Curve Fitting...' and perform a custom fit with IJ's built-in CurveFitter.
The CurveFitter has been recently revamped and while the documentation for the
new version is underway, you can have a look at the guide for IJ 1.45
<http://imagej.nih.gov/ij/docs/guide/userguide-27.htm>

But if you are already dealing with skeletons, did you consider using Ignatio's
Analyze Skeleton (in Fiji, Plugins>Skeleton>Analyze Skeleton),
<http://fiji.sc/wiki/index.php/AnalyzeSkeleton>?

> I would also like to utilize the radius span to minimize tangential
> intersection counts, but I'm not clear on how this works
See the documentation at
<http://imagejdocu.tudor.lu/doku.php?id=plugin:analysis:asa:start>
See also Tom's original documentation at
<http://www-biology.ucsd.edu/labs/ghosh/software/ShollAnalysis.pdf>

> Is it possible to save the graph of intersections/radii?  When I try to save
> it, it saves only the values, not the graph.
Use File>Save As>PNG... (or any other image format of your choice) to save the
canvas of any IJ plot.

If you think it helps, feel free to point us to a representative image

HTH,
-tiago

---

If you reply to this email, your message will be added to the discussion below:

http://imagej.1557.n6.nabble.com/Questions-regarding-Sholl-Analysis-output-tp4946474p4946766.html

To unsubscribe from Questions regarding Sholl Analysis output, click here.
NAML

[Tiago Ferreira-2]

Hi Jason,

On 2012.05.02, at 13:58 , jstanko wrote:
> I've attached a rat mammary gland whole mount, skeletonized image, Sholl mask, and output files for one of my samples.

I now understand the problem. You have been sampling the interstitial spaces around the gland. The plugin assumes that
the arbor has the same value as the center pixel. In your sample the center (964,1022) is outside the arbor and has a
value of 0. You should be getting this instead: <http://dl.dropbox.com/u/2172281/ShollMaskJason.png>.

I guess this could be considered a bug of the plugin. In IJ, the interpretation of binary images relies on the "Black
background" option in 'Process>Binary>Options...'. This is discussed, e.g., here:
<http://imagej.nih.gov/ij/docs/guide/userguide-26.html#infobox:blackBackground>
I'd have preferred to always interpret the arbor has having a value of 255, but ended up applying the strategy of using
the value of the focus to define it. In your case the focus is outside the arbor, so this obvious doesn't work. I will
try to solve this, but meanwhile, as a workaround, paint the pixel where you place the center mark with the arbor
intensity (the image you sent this would be black).


> What is the advantage of Ignatio's vs. Sholl?

The AnalyzeSkeleton is able to thoroughly describe topographic skeletons (2D and 3D). It would be complementary to your
analysis. It can measure parameters such as linearized length, # branches, largest shortest path, end-points. See the
plugin page for details, including the references on the bottom of page.


> I've read the Radius Span descriptions, I guess I'm not clear on how measurements are taken.

You won't be too much misguided if you consider it as the "thickness" of the sampling circle, in which you sample the
arbor with a donut-shape band that has the thickness of "Radius Span". Imagine you are sampling the letter "T". You
now specify the settings such as there is only a circle that intercepts the "T" right in its middle at radius r. In
classical Sholl, this would be one intersection (the vertical bar of the letter). With a span of, e.g., 2, you would
now draw the same circle, plus two others, one at r-1 and other at r+1. In this case you could obtain something like:
1 intersection at r, 1 intersection at r-1, 2 intersections at r+1 (the horizontal bar of the letter). You then
combine this 3 values (mean=1.3) to obtain a more accurate sample of the letter "T".

-tiago

[jstanko]

Hi Tiago,

 

Thanks again for your response.  I will review your suggestions and let you know if I have any additional questions.  Please let me know what you find in regards to starting outside the arbor and I will try the Ingatio’s method, it sounds like that may be more informative for my purposes than Sholl’s.

 

Jason

 

 

From: Tiago Ferreira-2 [via ImageJ] [mailto:[hidden email]]
Sent: Thursday, May 03, 2012 10:05 AM
To: Stanko, Jason (NIH/NIEHS) [E]
Subject: Re: Questions regarding Sholl Analysis output

 

Hi Jason,

On 2012.05.02, at 13:58 , jstanko wrote:
> I've attached a rat mammary gland whole mount, skeletonized image, Sholl mask, and output files for one of my samples.

I now understand the problem. You have been sampling the interstitial spaces around the gland. The plugin assumes that
the arbor has the same value as the center pixel. In your sample the center (964,1022) is outside the arbor and has a
value of 0. You should be getting this instead: <http://dl.dropbox.com/u/2172281/ShollMaskJason.png>.

I guess this could be considered a bug of the plugin. In IJ, the interpretation of binary images relies on the "Black
background" option in 'Process>Binary>Options...'. This is discussed, e.g., here:
<http://imagej.nih.gov/ij/docs/guide/userguide-26.html#infobox:blackBackground>
I'd have preferred to always interpret the arbor has having a value of 255, but ended up applying the strategy of using
the value of the focus to define it. In your case the focus is outside the arbor, so this obvious doesn't work. I will
try to solve this, but meanwhile, as a workaround, paint the pixel where you place the center mark with the arbor
intensity (the image you sent this would be black).


> What is the advantage of Ignatio's vs. Sholl?

The AnalyzeSkeleton is able to thoroughly describe topographic skeletons (2D and 3D). It would be complementary to your
analysis. It can measure parameters such as linearized length, # branches, largest shortest path, end-points. See the
plugin page for details, including the references on the bottom of page.


> I've read the Radius Span descriptions, I guess I'm not clear on how measurements are taken.

You won't be too much misguided if you consider it as the "thickness" of the sampling circle, in which you sample the
arbor with a donut-shape band that has the thickness of "Radius Span". Imagine you are sampling the letter "T". You
now specify the settings such as there is only a circle that intercepts the "T" right in its middle at radius r. In
classical Sholl, this would be one intersection (the vertical bar of the letter). With a span of, e.g., 2, you would
now draw the same circle, plus two others, one at r-1 and other at r+1. In this case you could obtain something like:
1 intersection at r, 1 intersection at r-1, 2 intersections at r+1 (the horizontal bar of the letter). You then
combine this 3 values (mean=1.3) to obtain a more accurate sample of the letter "T".

-tiago

---

If you reply to this email, your message will be added to the discussion below:

http://imagej.1557.n6.nabble.com/Questions-regarding-Sholl-Analysis-output-tp4946474p4949355.html

To unsubscribe from Questions regarding Sholl Analysis output, click here.
NAML

[jstanko]

Hi Tiago,

 

I downloaded the AnalyzeSkeleton plugin and have been trying to use it, but I can’t seem to find an adequate tutorial (there’s one for 3D, but I’m using 2D) so I have no idea if I’m doing it correctly or not.  Do you happen to have a link to detailed instructions?  The skeletonized image I’m using is essentially one big loop, so I’m getting 0 branches and obviously doing something wrong.  Also, for some reason I’m not seeing a progress bar.  I know that the analyses could take a while and I’m not sure if it’s finished or not.  If you like I can send you a screen shot of what I’m doing, maybe that would help.

 

Any progress on the workaround for the starting point pixel value?  I’ve been placing the focal point on a 255 pixel and getting the image you sent but I would like to be able to validate the data (somewhere verify that the intersection counts are correct on at least a few random rings).  However, the number of rings don’t correlate with the radial increments and data output.  I’m messing with the binary options to see if any of that helps.

 

Thanks,

Jason

 

From: Tiago Ferreira-2 [via ImageJ] [mailto:[hidden email]]
Sent: Thursday, May 03, 2012 10:05 AM
To: Stanko, Jason (NIH/NIEHS) [E]
Subject: Re: Questions regarding Sholl Analysis output

 

Hi Jason,

On 2012.05.02, at 13:58 , jstanko wrote:
> I've attached a rat mammary gland whole mount, skeletonized image, Sholl mask, and output files for one of my samples.

I now understand the problem. You have been sampling the interstitial spaces around the gland. The plugin assumes that
the arbor has the same value as the center pixel. In your sample the center (964,1022) is outside the arbor and has a
value of 0. You should be getting this instead: <http://dl.dropbox.com/u/2172281/ShollMaskJason.png>.

I guess this could be considered a bug of the plugin. In IJ, the interpretation of binary images relies on the "Black
background" option in 'Process>Binary>Options...'. This is discussed, e.g., here:
<http://imagej.nih.gov/ij/docs/guide/userguide-26.html#infobox:blackBackground>
I'd have preferred to always interpret the arbor has having a value of 255, but ended up applying the strategy of using
the value of the focus to define it. In your case the focus is outside the arbor, so this obvious doesn't work. I will
try to solve this, but meanwhile, as a workaround, paint the pixel where you place the center mark with the arbor
intensity (the image you sent this would be black).


> What is the advantage of Ignatio's vs. Sholl?

The AnalyzeSkeleton is able to thoroughly describe topographic skeletons (2D and 3D). It would be complementary to your
analysis. It can measure parameters such as linearized length, # branches, largest shortest path, end-points. See the
plugin page for details, including the references on the bottom of page.


> I've read the Radius Span descriptions, I guess I'm not clear on how measurements are taken.

You won't be too much misguided if you consider it as the "thickness" of the sampling circle, in which you sample the
arbor with a donut-shape band that has the thickness of "Radius Span". Imagine you are sampling the letter "T". You
now specify the settings such as there is only a circle that intercepts the "T" right in its middle at radius r. In
classical Sholl, this would be one intersection (the vertical bar of the letter). With a span of, e.g., 2, you would
now draw the same circle, plus two others, one at r-1 and other at r+1. In this case you could obtain something like:
1 intersection at r, 1 intersection at r-1, 2 intersections at r+1 (the horizontal bar of the letter). You then
combine this 3 values (mean=1.3) to obtain a more accurate sample of the letter "T".

-tiago

---

If you reply to this email, your message will be added to the discussion below:

http://imagej.1557.n6.nabble.com/Questions-regarding-Sholl-Analysis-output-tp4946474p4949355.html

To unsubscribe from Questions regarding Sholl Analysis output, click here.
NAML

[jstanko]

Hi Tiago,

 

Did you receive my last email?...

 

I downloaded the AnalyzeSkeleton plugin and have been trying to use it, but I can’t seem to find an adequate tutorial (there’s one for 3D, but I’m using 2D) so I have no idea if I’m doing it correctly or not.  Do you happen to have a link to detailed instructions?  The skeletonized image I’m using is essentially one big loop, so I’m getting 0 branches and obviously doing something wrong.  Also, for some reason I’m not seeing a progress bar.  I know that the analyses could take a while and I’m not sure if it’s finished or not.  If you like I can send you a screen shot of what I’m doing, maybe that would help.

 

Any progress on the workaround for the starting point pixel value?  I’ve been placing the focal point on a 255 pixel and getting the image you sent but I would like to be able to validate the data (somewhere verify that the intersection counts are correct on at least a few random rings).  However, the number of rings don’t correlate with the radial increments and data output.  I’m messing with the binary options to see if any of that helps.

 

Thanks,

Jason

 

 

From: Tiago Ferreira-2 [via ImageJ] [mailto:[hidden email]]
Sent: Thursday, May 03, 2012 10:05 AM
To: Stanko, Jason (NIH/NIEHS) [E]
Subject: Re: Questions regarding Sholl Analysis output

 

Hi Jason,

On 2012.05.02, at 13:58 , jstanko wrote:
> I've attached a rat mammary gland whole mount, skeletonized image, Sholl mask, and output files for one of my samples.

I now understand the problem. You have been sampling the interstitial spaces around the gland. The plugin assumes that
the arbor has the same value as the center pixel. In your sample the center (964,1022) is outside the arbor and has a
value of 0. You should be getting this instead: <http://dl.dropbox.com/u/2172281/ShollMaskJason.png>.

I guess this could be considered a bug of the plugin. In IJ, the interpretation of binary images relies on the "Black
background" option in 'Process>Binary>Options...'. This is discussed, e.g., here:
<http://imagej.nih.gov/ij/docs/guide/userguide-26.html#infobox:blackBackground>
I'd have preferred to always interpret the arbor has having a value of 255, but ended up applying the strategy of using
the value of the focus to define it. In your case the focus is outside the arbor, so this obvious doesn't work. I will
try to solve this, but meanwhile, as a workaround, paint the pixel where you place the center mark with the arbor
intensity (the image you sent this would be black).


> What is the advantage of Ignatio's vs. Sholl?

The AnalyzeSkeleton is able to thoroughly describe topographic skeletons (2D and 3D). It would be complementary to your
analysis. It can measure parameters such as linearized length, # branches, largest shortest path, end-points. See the
plugin page for details, including the references on the bottom of page.


> I've read the Radius Span descriptions, I guess I'm not clear on how measurements are taken.

You won't be too much misguided if you consider it as the "thickness" of the sampling circle, in which you sample the
arbor with a donut-shape band that has the thickness of "Radius Span". Imagine you are sampling the letter "T". You
now specify the settings such as there is only a circle that intercepts the "T" right in its middle at radius r. In
classical Sholl, this would be one intersection (the vertical bar of the letter). With a span of, e.g., 2, you would
now draw the same circle, plus two others, one at r-1 and other at r+1. In this case you could obtain something like:
1 intersection at r, 1 intersection at r-1, 2 intersections at r+1 (the horizontal bar of the letter). You then
combine this 3 values (mean=1.3) to obtain a more accurate sample of the letter "T".

-tiago

---

If you reply to this email, your message will be added to the discussion below:

http://imagej.1557.n6.nabble.com/Questions-regarding-Sholl-Analysis-output-tp4946474p4949355.html

To unsubscribe from Questions regarding Sholl Analysis output, click here.
NAML