Posted by
Tiago Ferreira-2 on
Sep 09, 2014; 4:51pm
URL: http://imagej.273.s1.nabble.com/Scholl-analysis-Fiji-tp5009555p5009560.html
Hi Chiara,
On Sep 9, 2014, at 09:33, Valori, Chiara /DZNE <
[hidden email]> wrote:
> I'm performing Scholl analysis with the Fiji plugin "simple neurite tracer".
Note that while Scholl is a common surname, it is irrelevant here.
The eponymous technique you are performing is named after Donald Sholl, the first
anatomist able to quantify the anatomy of neuronal arbors in single-cell labeling,
more than 60 years ago (See PMID 13117757 in
http://fiji.sc/Sholl#References).
> I have a question about the Sholl regression coefficient which is calculated in
> the Results table: I do not understand what this number represents when cells are
> analyzed using standard axes.
In short: It means nothing when using linear axes, as it only applies to semi-log
or log-log transformations, as you suspected. So, why does SNT, reports it? I'm
guessing there were two reasons why Mark decided to report it at all times:
1) SNT features an extremely useful interactive prompt that automatically updates
calculations every time a checkbox gets toggled. The calculations are performed
in the background even when the user has chosen a combination of choices that
do not make much sense. The plugin does not try to outsmart the user (wouldn't
it be annoying otherwise?[1]) and just reports whatever calculations it was
asked to.
2) Historically, ImageJ did not allow for non-mumeric text in the Results Table
(this is no longer the case with recent versions of ImageJ). So including a
non-numeric value for the Regression coefficient when using linear axes in the
Results table, was not a feasible option.
That being said, we probably could change the plugin if you think this is a bug.
Note however that using log axes but choosing "No normalization" would not make
sense either. Hopefully this is all better explained throughout the
http://fiji.sc/Sholl documentation page (mainly
http://fiji.sc/Sholl#MethodsTable).
> I thought you need a polynomial function for standard and thus more parameters...
You're also right. Linear axes (i.e, no normalizations to Area or Volume) would
require a polynomial fit. For that you would need to use Analyze>Sholl Analysis,
as indicated in
http://fiji.sc/Sholl#ImportingThis would also give you access to a wide repertoire of metrics, not available in
SNT.
Best,
-tiago
[1] In Sholl Analysis (Analyze>Sholl Analysis,
http://fiji.sc/Sholl), the plugin
does try to outsmart the user, and will automatically disable options that do
not fulfill critical requirements. The consequence however, is that the prompt
is not as streamlined as in SNT.
--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html