volume integration under 3d surface

[Tommaso Nicolini]

Hi everyone,
I'm kind of new to imageJ since I'm a chemist and just started working on
Electron Crystallography.
I'm analyzing my electron diffraction patterns (tiff file) with imageJ
because crystallography software is either too expensive or developed for
inorganic materials or proteines and most common methods don't apply to
polymeric materials.
The data I need to gather from my files are the actual intensities of the
reflections and to do this i would need to calculate the volume under each
peak of grayscale values of my 8-bit image file. Has anybody ever done this?
(I don't have matlab at the moment but from what i read it might be my only
option but the problem then would be how to convert the tiff file into an
actual matrix of numerical values)

Any type of suggestion would be greatly appreciated!


Tommaso

[reinhard rachel]

>>>
> ... Electron Crystallography.
> I'm analyzing my electron diffraction patterns (tiff file) with imageJ
> because crystallography software is either too expensive or developed for
> inorganic materials or proteines and most common methods don't apply to
> polymeric materials.
> The data I need to gather from my files are the actual intensities of the
> reflections and to do this i would need to calculate the volume under each
> peak of grayscale values of my 8-bit image file. Has anybody ever done this?
> (I don't have matlab at the moment but from what i read it might be my only
> option but the problem then would be how to convert the tiff file into an
> actual matrix of numerical values)

Dear Tommaso,
I am a bit worried that you try to analyze data which are - according to your mail - electron diffraction patterns, but, they are 8-bit only. To the best of my knowledge, electron diffraction data usually have a huge dynamic range, by far exceeding 8bit depth - so you already might have lost valuable and important scientific information, due to the conversion to 8bit. I am not quite sure whether it is worth to analyze these data further ... whatever you do.
In addition, I do not see how the electrons you are "using" would be able to differentiate between polymer A (protein) and polymer B (your chemical polymer).
Sorry that I cannot help any further, but some of the electron diffraction specialists might be able?
kind regards,
Reinhard


--
Prof. Dr. Reinhard Rachel
University of Regensburg
Centre for EM / Anatomy
Faculty of Biology & Preclin. Med.
Universitaetsstrasse 31
D-93053 Regensburg - Germany
tel [phone], 1720
fax [phone]
mail [hidden email]
office: VKL 3.1.29
 
next microscopy conferences:
http://www.emc2012.org.uk/
EMC 2012 in Manchester, UK
http://www.mc2013.de/
MC2013 in Regensburg, Germany

[Michael Schmid]

Hi Tommaso,

for the integrating over a spot, you can simply use 'Measure' with 'Integrated density' selected in the Measurement options.
See the documentation, http://rsb.info.nih.gov/ij/docs/guide/userguide-27.html#sub:Set-Measurements...

Before this, you have to subtract the background. The easiest would probably be converting your image to float (Image>Type>32 bit) and somehow determining the average value of the background (e.g. 'mode' in the histogram), and use Process>Math>Subtract to subtract this. If you have a sloping or uneven background, you may run 'Subtract Background' before this (e.g. with 'sliding paraboloid' and a rather large radius; use 'preview' with 'create background' to make sure that your diffraction spots do not appear in the background).

I agree with Reinhard Rachel that you should get better data with better resolution than 8 bit.


Michael
________________________________________________________________
On Apr 11, 2012, at 19:29, Tommaso Nicolini wrote:

> Hi everyone,
> I'm kind of new to imageJ since I'm a chemist and just started working on
> Electron Crystallography.
> I'm analyzing my electron diffraction patterns (tiff file) with imageJ
> because crystallography software is either too expensive or developed for
> inorganic materials or proteines and most common methods don't apply to
> polymeric materials.
> The data I need to gather from my files are the actual intensities of the
> reflections and to do this i would need to calculate the volume under each
> peak of grayscale values of my 8-bit image file. Has anybody ever done this?
> (I don't have matlab at the moment but from what i read it might be my only
> option but the problem then would be how to convert the tiff file into an
> actual matrix of numerical values)
>
> Any type of suggestion would be greatly appreciated!
>
>
> Tommaso

[Tommaso Nicolini]

Thank you all for your prompt suggestions and considerations.
I have already been working with background subtraction with imageJ and
also other programs for diffraction and the best procedure I found was with
a roving window method in fibrefix (software for treatment of fiber
diffraction patterns).
After background subtraction the patterns look satisfactorily flat
(checking the profile with imageJ) and the intensities look unaffected by
comparison with the subtracted background. Now thanks to Michael Schmid's
tip I'm able to measure the intensities and I hope to check for
reproducibility of collected data from various patterns.
By the way I probably  wasn't clear in saying that usual diff software
isn't appropriate for my patterns: this is because they usually employ
fitting procedures to measure the intensities and I have just 15-20
measurable reflections as opposed to typical protein diffraction patterns
which present a larger number of reflections (also because of different
size of the unit cell) and this surely represents a limit for the
information that can be extracted from my data. I also encountered a
problem with fibrefix when trying to impose the appropriate spacegroup
during the fitting procedure so we opted for this approximated method for
extracting intensities hoping it will yield some results.
If preliminary results seem ok I might try to get access to a TEM equipped
with a better CCD camera because the one I'm using to learn TEM isn't set
for getting images other than 8-bit at the moment, but I made sure
reflections weren't burnt out so I should be ok.

Thanks again for the help

Tommaso