Posted by
karo03 on
URL: http://imagej.273.s1.nabble.com/Quantifying-Stained-Retina-tp5000880p5001010.html
Ok, the first step you did. Clearly the integrated density reflects the region size.
Hence what is the related area to get a reliable estimator of mean density. I would recommend either the cells, if they are segmentable, or the inverse of the white background, e.g. taken from the blue channel. This makes sense for regions, where the cells are recognizable.
The regions of each layer should be defined as large as possible (area=A). To correct a bit the areas, measure also the white area (background, BG) inside the regions.
Than calculate: mean density=integrated density / (A-BG) per region.
I don't know how ImageJ calculates integrated density, I would recommend to transform the channel image to "extinction" by
run("32-bit");
run("Divide...", "value=255");
# 255 is the white incoming light, possibly it is a bit less.
# The pixel values of your image are the transmitted light.
run("Log");
run("Multiply...", "value=-1");
//run("Brightness/Contrast..."); # to make the image content visible
and measure than the mean gray value. Than is integrated density = mean gray value * (A) and mean density can than be calculated as given above. Following the infos from Joel it makes sense to do these measurements on each channel! Possibly also a color deconvolution (PCA ?) could make sense to discriminate the cell stain from the iron stain.
If this/these "mean density" does/do not differ enough, meaning that the amount of iron per layer is not differentiable with these measures, the next step starts: Up to now it is assumed, that the section thickness is constant and the stain reaches homogeneously the iron in the whole material. This has to be proven, e.g. by testing some samples (not the delivered images) from somebody with expertise in (quantitative) microscopy! Critical is the section thickness!
I think with the numbers calculated as described for a senseful set of images (with layers) per group (?), it will be not completely easy to state "non-discriminability" of iron content!
Best regards
Karsten
Am 29.11.2012 um 21:05 schrieb Natalia Chacon:
> Dear Karsten,
>
> How would you recommend measuring these ROI?
>
> I've tested out the basic Analyze>Set Measurements, Analyze>
> Measurement on a Split Channel (RGB). With this process I measured the
> area,integrated density, and mean gray density of the Split Channel
> Blue image.
>
> The issue with this analysis is that there is no true distiction in
> the amount of "Blue"
> Rather, all of the values ended up to be the total amount of pixels
> that compose the image.
>
> Natalia
>
> On 11/28/12, Karsten <
[hidden email]> wrote:
>> Hi,
>> this is an interesting thread with many very differing layers of
>> understanding of images!
>>
>> What about measuring the "amount" of blue, which is IMHU the total
>> extinction of the blue channel, in a selected region, e.g. a layer or a part
>> of it? For standardization the area of the region should be measured too, or
>> if the nuclei are of interest, better the area of the nuclei inside the
>> selected region.
>>
>> ... and so on, step by step a real measurement scheme will be established.
>>
>> For reliable measurements lots of questions will arise! If "the iron"
>> stained by "Pearls Blue" is in any sense stoichiometric perhaps images
>> should be sampled as pure intensity image using the appropriate filter!
>>
>> However, I would recommend to measure, as the mentor ask for, and look at
>> the results.
>> Also mentors are still learning!
>>
>> Regards
>> Karsten
>>
>> Am 28.11.2012 um 21:55 schrieb Natalia Chacon:
>>
>>> On 11/28/12, Natalia Chacon <
[hidden email]> wrote:
>>>> Dear Mr. Gabriel,
>>>>
>>>> I've enclosed another image that seems to contain more iron in the
>>>> sample.
>>>> Also, is it still possible to use the Color Deconvolution to get the
>>>> data my mentor desires?
>>>>
>>>> Thanks,
>>>> Natalia
>>>>
>>>> On 11/20/12, Gabriel Landini <
[hidden email]> wrote:
>>>>> On Tuesday 20 Nov 2012 14:55:39 Natalia Chacon wrote:
>>>>>> Rob, my mentor told me she used a stain called "Pearls Blue"
>>>>>>
>>>>>> Joel, I'm interested in quantifiying most of the stainings visible.
>>>>>> (the background staining I don't desire to quantify.)
>>>>>>
>>>>>> What you're stating though does correlate with what my mentor has
>>>>>> informed
>>>>>> me. That the onl nuclei would be the most dense. This would be due to
>>>>>> the
>>>>>> way the protien settles.
>>>>>
>>>>> This is what Prussian, Pearls or Berlin blue looks like:
>>>>>
>>>>>
http://www.jichi.ac.jp/pathology/swfu/d/fe-10.jpeg>>>>>
>>>>> Your image looks like all contrast stain to me. So you might not have
>>>>> any
>>>>> iron
>>>>> compounds there or it is so faint that the contrast stain prevails.
>>>>>
>>>>> With Berlin blue, you can *demonstrate* that some iron compound is
>>>>> there,
>>>>> but
>>>>> I am not sure if it can be used to quantify their quantity. Here it
>>>>> says
>>>>> it
>>>>>
>>>>> cannot, so I would be careful how to interpret the intensity:
>>>>>
>>>>>
http://stainsfile.info/StainsFile/stain/pigment/perls.htm>>>>>
>>>>> Regards
>>>>>
>>>>> Gabriel
>>>>>
>>>>> --
>>>>> ImageJ mailing list:
http://imagej.nih.gov/ij/list.html>>>>>
>>>>
>>>
>>> --
>>> ImageJ mailing list:
http://imagej.nih.gov/ij/list.html>>> <BB.JPG>
>>
>> --
>> ImageJ mailing list:
http://imagej.nih.gov/ij/list.html>>
>
> --
> ImageJ mailing list:
http://imagej.nih.gov/ij/list.htmlKarsten
[hidden email]
--
ImageJ mailing list:
http://imagej.nih.gov/ij/list.html