areatree results: LB cable length

[Luisa de Vivo]

Hi,

does anybody know the difference between Cable length and LB Cable length
in the Areatree results window of TrakEM2?
Thank you very much,

Luisa


--
Luisa de Vivo, PhD
Department of Psychiatry
University of Wisconsin
6001 Research Park Boulevard
Madison WI 53719

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

[Albert Cardona-2]

The Lower Bound (LB) cable length is a way to approximate the minimal
possible cable length that the tree could have. Consists of measuring as
straight lines the distance between end node and branch node, branch node
to branch node, and branch node to root, and adding them all up.

The main effect of the LB measurement is that the jitter from the human's
innaccuracy and the jitter from the image registration errors are almost
completely removed. This is particularly important for EM reconstructions
from serial sections; if you use light-microscopy image stacks the jitter
cannot be much and is likely not significant.

There is an example of this measurement in supplemental figure 15, and an a
schematic of its measurement in suppl. figure 16 of this paper:

http://www.nature.com/nmeth/journal/v9/n7/full/nmeth.2072.html

"Elastic volume reconstruction from series of ultra-thin microscopy
sections"
Stephan Saalfeld, Richard Fetter, Albert Cardona & Pavel Tomancak
Nature Methods 9, 717–720 (2012) doi:10.1038/nmeth.2072

In there, we demonstrate how better image registration brings the cable
length of a hand-reconstructed neuronal arbor closer to the lower bound
measurement. The latter stays nearly constant for all image registration
techniques used, which endorses its robustness.

Albert



2014/1/20 Luisa de Vivo <[hidden email]>

> Hi,
>
> does anybody know the difference between Cable length and LB Cable length
> in the Areatree results window of TrakEM2?
> Thank you very much,
>
> Luisa
>
>
> --
> Luisa de Vivo, PhD
> Department of Psychiatry
> University of Wisconsin
> 6001 Research Park Boulevard
> Madison WI 53719
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
http://albert.rierol.net
http://www.ini.uzh.ch/~acardona/

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

[Albert Cardona-2]

> On Jan 21, 2014, at 6:34 PM, Luisa de Vivo <[hidden email]> wrote:
>
> Thanks. I don't know whether I should use the LB measurement though: on one hand it removes the "artificial" jitters but on the other hand does not ignore the natural winding of dendrites that can slightly bend and turn in the tissue? In some of my dendrites, the cable length and the LB cable length are very different (LB half the cable length)... do you think it is always better to choose the LB measurement? I need to know the length of the dendritic shaft in order to calculate spine density.
>
> Thank you very much,
>
> Luisa


The LB measurement can be too low if you neurons don't branch a lot. It is there to provide a reference. A proper measurement would be the cable length of a tree smoothed by convolving with a gaussian, or as a reasonable approximation, a running three-point average that would remove most jitter. The latter, if not available in TrakEM2 (I thought it was, as a "smooth" measurement?), would be quite trivial to compute with a script.

Albert

>
>
>> On Mon, Jan 20, 2014 at 6:35 PM, Albert Cardona <[hidden email]> wrote:
>> The Lower Bound (LB) cable length is a way to approximate the minimal possible cable length that the tree could have. Consists of measuring as straight lines the distance between end node and branch node, branch node to branch node, and branch node to root, and adding them all up.
>>
>> The main effect of the LB measurement is that the jitter from the human's innaccuracy and the jitter from the image registration errors are almost completely removed. This is particularly important for EM reconstructions from serial sections; if you use light-microscopy image stacks the jitter cannot be much and is likely not significant.
>>
>> There is an example of this measurement in supplemental figure 15, and an a schematic of its measurement in suppl. figure 16 of this paper:
>>
>> http://www.nature.com/nmeth/journal/v9/n7/full/nmeth.2072.html
>>
>> "Elastic volume reconstruction from series of ultra-thin microscopy sections"
>> Stephan Saalfeld, Richard Fetter, Albert Cardona & Pavel Tomancak
>> Nature Methods 9, 717–720 (2012) doi:10.1038/nmeth.2072
>>
>> In there, we demonstrate how better image registration brings the cable length of a hand-reconstructed neuronal arbor closer to the lower bound measurement. The latter stays nearly constant for all image registration techniques used, which endorses its robustness.
>>
>> Albert
>>
>>
>>
>> 2014/1/20 Luisa de Vivo <[hidden email]>
>>> Hi,
>>>
>>> does anybody know the difference between Cable length and LB Cable length
>>> in the Areatree results window of TrakEM2?
>>> Thank you very much,
>>>
>>> Luisa
>>>
>>>
>>> --
>>> Luisa de Vivo, PhD
>>> Department of Psychiatry
>>> University of Wisconsin
>>> 6001 Research Park Boulevard
>>> Madison WI 53719
>>>
>>> --
>>> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>>
>>
>>
>> --
>> http://albert.rierol.net
>> http://www.ini.uzh.ch/~acardona/
>
>
>
> --
> Luisa de Vivo, PhD
> Department of Psychiatry
> University of Wisconsin
> 6001 Research Park Boulevard
> Madison WI 53719
>

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

[jrybak]

I would like to measure the neuron length directly from the area list, that I reconstructed from serial EM sections.
Is that possible ?
I tried to export the area list as an Amira file. Here, I could convert the surface to a Skeleton.  but the file is label file too large, even with downsampling the export function uses (does she ?).

Is there a way the area list, i.e the reconstructed neuron to convert to a construct, like a lineset, or Skeleton
in order to read out parameters like neuron length, or make a dendrogramm ??

Many thanks
Jürgen

 

[Albert Cardona-2]

2014-06-30 12:18 GMT-04:00 jrybak <[hidden email]>:

> I would like to measure the neuron length directly from the area list,
> that I
> reconstructed from serial EM sections.
> Is that possible ?
> I tried to export the area list as an Amira file. Here, I could convert the
> surface to a Skeleton.  but the file is label file too large, even with
> downsampling the export function uses (does she ?).
>
> Is there a way the area list, i.e the reconstructed neuron to convert to a
> construct, like a lineset, or Skeleton
> in order to read out parameters like neuron length, or make a dendrogramm
> ??
>
> Many thanks
> Jürgen
>
>
>

Jurgen,

to avoid creating an enormous image stack, one way to go about it is to
create an ImgLib image that uses a list of java.awt.Shape objects as the
data, rather than pixels. This is the approach taken for e.g. running
marching cubes on an AreaList [1]. With that ImgLib image, you could then
write a program to skeletonize it, which would be a generic ImgLib
algorithm--maybe it already exists, ask the ImgLib developers first.

An alternative that doesn't involve any code is to generate the labels
stack as a series of 2D images, and then open them as a virtual stack. The
performance of the skeletonizer will likely be poor, but it will do the job.

Best,

Albert

[1]
https://github.com/fiji/TrakEM2/blob/master/TrakEM2_/src/main/java/ini/trakem2/utils/AreaUtils.java#L124

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