After using the GelAnalyzer plugin for some time, I noticed that
depending on the selection sizes and screen settings ImageJ was
compressing the output plots. Since using this plugin for
quantitation relies directly on these plots and not the original
image, I find this concerning.
Digging through the source
(
http://rsb.info.nih.gov/ij/docs/source/ij/plugin/GelAnalyzer.java.html),
I found this:
if(isVertical)
plotWidth = firstRect.height;
else
plotWidth = firstRect.width;
if (plotWidth<500)
plotWidth = 500;
if(isVertical) {
if (plotWidth>2*firstRect.height)
plotWidth = 2*firstRect.height;
} else {
if (plotWidth>2*firstRect.width)
plotWidth = 2*firstRect.width;
}
Dimension screen = Toolkit.getDefaultToolkit().getScreenSize();
if (plotWidth>screen.width-40)
plotWidth = screen.width - 40;
plotHeight = plotWidth/2;
if (plotHeight<200)
plotHeight = 200;
if (plotHeight>400)
plotHeight = 400;
Considering that the intrinsic error of an area defined by a pixelated
image is basically the area of the perimeter divided by the area of
the enclosed region, isn't it in our interest to at least keep the
plots true to the original data and not compress them? Some testing
with my experimental images indicates that these intrinsic errors can
be significant.
I've rewritten the GelAnalyzer code to deal with this issue, but I
would like to get input from people who have more experience than I
with the technical issues of image densitometry. It seems like the
best solution is to maximize the final plot size with emphasis on the
horizontal axis, since maintaining the original ratio of
signal:(selection length) would give extremely tall plots, and thus
very non-circular areas, which increase the perimeter/area ratios.
As a related issue, using code that generates plots > 5000x5000 pixels
causes artifacts on a Windows system with the most recent JDK and
ImageJ release: I'm assuming this is related to the size of the
images. Are there any suggestions or fixes for this issue?
Jonathan Hilmer
Bothner Research Group
Department of Chemistry and Biochemistry
Montana State University
[hidden email]