correction for stray light on epifluorescence images

[Stoyan Pavlov]

Dear all,
I've got epifluorescence images, on which I am to quantify the total
fluorescence (integrated density) of ROIs. On some of the images there are
bright glowing artifacts. I exclude them from measurement however the stray
light from them affects the neighbouring pixels. So i am searching for a
way to neutralize this influence.
My images are corrected for illumination irregularities (shading
correction) at the acquisition.
I was wondering would it be legitimate to do this correction dividing all
my images by their smoothed with a big median filter (>100px) versions. I
would do this using the Image calculator plus divide formula (i1/12)*k1
where k1 would be the maximum grey value of the smoothed version.

Best regards!

Stoyan Pavlov
---
Dr. Stoyan P. Pavlov, MD, PhD
Departament of Anatomy, Histology and Embryology
Medical University "Prof. Dr. Paraskev Stoyanov", Varna
Prof. Marin Drinov Str.55
9002 Varna
 Bulgaria
Tel: +359 (0) 52 - 677 - 052
e-mail: [hidden email]

Zentrum für Anatomie
Uniklinik Köln
[hidden email]
Tel: [phone]

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

[karo03]

Hi,
I think your idea with median is feasible, provided your objects of interest are small enough to be smoothed away and the artifacts are large enough not to be smoothed away.

However, why not using "Background Subtract"? If you think division is better, than logarithm the image before. Be careful in this case with resulting values less than 1!

Still I have a problem of understanding: You say you perform shading correction! In fluorescence? with which background image? Illumination is the excitation light, if there is any background, I think you should better inspect your microscope and improve the settings. Even the amount of stray light could be reasonably reduced!

Regards
Karsten

Am 16.10.2013 um 09:10 schrieb Stoyan Pavlov <[hidden email]>:

> Dear all,
> I've got epifluorescence images, on which I am to quantify the total
> fluorescence (integrated density) of ROIs. On some of the images there are
> bright glowing artifacts. I exclude them from measurement however the stray
> light from them affects the neighbouring pixels. So i am searching for a
> way to neutralize this influence.
> My images are corrected for illumination irregularities (shading
> correction) at the acquisition.
> I was wondering would it be legitimate to do this correction dividing all
> my images by their smoothed with a big median filter (>100px) versions. I
> would do this using the Image calculator plus divide formula (i1/12)*k1
> where k1 would be the maximum grey value of the smoothed version.
>
> Best regards!
>
> Stoyan Pavlov
> ---
> Dr. Stoyan P. Pavlov, MD, PhD
> Departament of Anatomy, Histology and Embryology
> Medical University "Prof. Dr. Paraskev Stoyanov", Varna
> Prof. Marin Drinov Str.55
> 9002 Varna
> Bulgaria
> Tel: +359 (0) 52 - 677 - 052
> e-mail: [hidden email]
>
> Zentrum für Anatomie
> Uniklinik Köln
> [hidden email]
> Tel: [phone]
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

Karsten
[hidden email]

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

[Stoyan Pavlov]

Hi Karsten,
As I use a small magnification objective so my  sources of positive
fluorescence are below or at least at the resolving power of the microscope
(so they might be considered as point sources).I do not think that in this
case Background subtraction can be useful. I actually do not need to remove
the artifacts completely but to reduce their influence on the brightness of
their neighbourhood. Usually when I am counting objects and measuring areas
I use Top-Hat Opening by Reconstruction to deal with artifacts and
irregularities, but i think in this case this would destroy the intensity
information in the images and render the results incomparable.
As to the shading correction: I do not have a background fluorescence. I
perform the shading correction to correct for the irregularities in the
illumination light - although my lamp is perfectly aligned there is still
unequal distribution of the incident exciting light resulting in difference
in the emission as well. To perform the shading correction I acquired a​
background image of an uniformly fluorescing  preparation (thin film of my
detector  - IgG-Cy3 conjugate) and saved it as a flat-field correction
image in the camera software. It works perfectly - all lamp induced
gradients disappear from my images.

Why is the logarithm prior the division necessary and what will it achieve?

Best regards
Stoyan

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

[karo03]

Hi Pavlov,
thank you for the explanations.

Logarithm before subtraction means division!

Interesting your shading correction. Perfectly aligned light is promising, did you however tea to Köhler the excitation light? That should be easily possible with your reference specimen! How is your shading correction done? by subtraction or by division? Your next step would mean a second correction which might make things somewhat complicated in understanding the complete system!

If you have such small objects I still think that top hat with or without reconstruction is best choice, of course without theresholding. The resulting peaks reflect IMHO good intensity signals for measuring. However don't forget that differing excitations are not corrected. The fluorescence signal is by no means linearly related to the excitation signal strength!

Thank you to let me share your considerations.

Regards

Karsten

Am 16.10.2013 um 11:15 schrieb Stoyan Pavlov <[hidden email]>:

> Hi Karsten,
> As I use a small magnification objective so my  sources of positive
> fluorescence are below or at least at the resolving power of the microscope
> (so they might be considered as point sources).I do not think that in this
> case Background subtraction can be useful. I actually do not need to remove
> the artifacts completely but to reduce their influence on the brightness of
> their neighbourhood. Usually when I am counting objects and measuring areas
> I use Top-Hat Opening by Reconstruction to deal with artifacts and
> irregularities, but i think in this case this would destroy the intensity
> information in the images and render the results incomparable.
> As to the shading correction: I do not have a background fluorescence. I
> perform the shading correction to correct for the irregularities in the
> illumination light - although my lamp is perfectly aligned there is still
> unequal distribution of the incident exciting light resulting in difference
> in the emission as well. To perform the shading correction I acquired a​
> background image of an uniformly fluorescing  preparation (thin film of my
> detector  - IgG-Cy3 conjugate) and saved it as a flat-field correction
> image in the camera software. It works perfectly - all lamp induced
> gradients disappear from my images.
>
> Why is the logarithm prior the division necessary and what will it achieve?
>
> Best regards
> Stoyan
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

Karsten
[hidden email]

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

[Stoyan Pavlov]

Thank you again , Karsten!
I misread "Logarithm before division" instead "Logarithm before
subtraction", hence my question :)
Of course you are right that fluorescence signal isn't a linear response to
excitation  but still difference in the power of incident light would
result in differences in excitation. And since I am not measuring peaks,
but I am interested in the total amount of light that is emitted from the
sampled volume, I guess the error will be small enough.
I will stay in my case away from the morfological filtering because it will
remove unknown to me amount of the out of focus light. My preparations are
pretty thick and i do not know how much the different sections shrinked.
That is why I want to collect and use for comparison as much as posible of
the light emitted from the samples (hence my choise of an objective of
small aperture and wide depth of field), and since the initial volumes were
the same, I guess that my results in terms of integrated densities will be
comparable, while the introduced bias will be small and relatively constant
amongst the different images.
I do not know the exact algorithm for the shading correction since it is a
part of the camera software, but judging by some tests I made and the
similarities between images I corrected manually by division and images
corrected by the Leica software I think that it is some form of a
correction by division.
Regards

Stoyan

---
Dr. Stoyan P. Pavlov, MD, PhD
Departament of Anatomy, Histology and Embryology
Medical University "Prof. Dr. Paraskev Stoyanov", Varna
Prof. Marin Drinov Str.55
9002 Varna
 Bulgaria
Tel: +359 (0) 52 - 677 - 052
e-mail: [hidden email]

Zentrum für Anatomie
Uniklinik Köln
[hidden email]
Tel: [phone]





2013/10/16 Karsten <[hidden email]>

> Hi Pavlov,
> thank you for the explanations.
>
> Logarithm before subtraction means division!
>
> Interesting your shading correction. Perfectly aligned light is promising,
> did you however tea to Köhler the excitation light? That should be easily
> possible with your reference specimen! How is your shading correction done?
> by subtraction or by division? Your next step would mean a second
> correction which might make things somewhat complicated in understanding
> the complete system!
>
> If you have such small objects I still think that top hat with or without
> reconstruction is best choice, of course without theresholding. The
> resulting peaks reflect IMHO good intensity signals for measuring. However
> don't forget that differing excitations are not corrected. The fluorescence
> signal is by no means linearly related to the excitation signal strength!
>
> Thank you to let me share your considerations.
>
> Regards
>
> Karsten
>
> Am 16.10.2013 um 11:15 schrieb Stoyan Pavlov <[hidden email]>:
>
> > Hi Karsten,
> > As I use a small magnification objective so my  sources of positive
> > fluorescence are below or at least at the resolving power of the
> microscope
> > (so they might be considered as point sources).I do not think that in
> this
> > case Background subtraction can be useful. I actually do not need to
> remove
> > the artifacts completely but to reduce their influence on the brightness
> of
> > their neighbourhood. Usually when I am counting objects and measuring
> areas
> > I use Top-Hat Opening by Reconstruction to deal with artifacts and
> > irregularities, but i think in this case this would destroy the intensity
> > information in the images and render the results incomparable.
> > As to the shading correction: I do not have a background fluorescence. I
> > perform the shading correction to correct for the irregularities in the
> > illumination light - although my lamp is perfectly aligned there is still
> > unequal distribution of the incident exciting light resulting in
> difference
> > in the emission as well. To perform the shading correction I acquired a​
> > background image of an uniformly fluorescing  preparation (thin film of
> my
> > detector  - IgG-Cy3 conjugate) and saved it as a flat-field correction
> > image in the camera software. It works perfectly - all lamp induced
> > gradients disappear from my images.
> >
> > Why is the logarithm prior the division necessary and what will it
> achieve?
> >
> > Best regards
> > Stoyan
> >
> > --
> > ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>
> Karsten
> [hidden email]
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

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