Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
Color Deconvolution Imagej Download
69 views
Skip to first unread message
Ronald Raynoso
Jan 9, 2024, 1:04:54 AM1/9/24
to
In resume, while colour deconvolution can be useful to segment immunostained structures (e.g. count cells, nuclei, identify structures) or for image enhancement (e.g. for colour blind observers), attempting to quantify DAB intensity as a measure of antigen expression using this plugin is not a good idea.
color deconvolution imagej download
Download https://1niosetincto.blogspot.com/?pu=2x5Ybs
3. Run the Colour Deconvolution2 plugin and select the From ROI option to define example stained areas. Click OK on the plugin dialogue.
The plugin will then ask for rectangular selections of single dye regions. Select ROIs on areas which are all intensely stained (saturated) with only one dye, without empty background regions. Repeat this for other dyes.
If the staining method consists of only two dyes, when prompted for the 3rd selection just right-click. That vector will then be set to 0,0,0 and computed according to the option set for Cross product for Colour 3 as discussed above.
The deconvolution of the test image will take place and the stain separation can be evaluated. The Log window will show something like this (with different values, of course):
Invalid citation arguments: "last"=>"Ruifrok", "first"=>"A.C.", "last2"=>"Johnston", "first2"=>"D.A.", "title"=>"Quantification of histochemical staining by color deconvolution", "journal"=>"Anal. Quant. Cytol. Histol.", "volume"=>"23", "pages"=>"291-299", "year"=>"2001", "PMID"=>"11531144"
Hi,
I uploaded a new version (2.1) of the colour deconvolution2 plugin.
Instructions for ImageJ and Fiji (as well as how to uninstall the one that comes by default with Fiji) can be found here:
-landini-software/colour-deconvolution-2/
Hello! Any tips on using colour deconvolution for analyzing Sirius Red Stains? Or if anyone has a good method for analyzing collagen area from Sirius Red stains in general, that would also be helpful. Thank you!
This version of the Colour Deconvolution plugin works for a single RGB image at the time and only allows a 2-colour deconvolution. Issues in this plugin at this point are that it does not put color...
Colour deconvolution is a method used in diagnostic brightfield microscopy to transform colour images of multiple stained biological samples into images representing the stain concentrations. It is applied by decomposing the absorbance values of stain mixtures into absorbance values of single stains. The method assumes a linear relation between stain concentration and absorbance, which is only valid under monochromatic conditions. Diagnostic applications, in turn, are often performed under polychromatic conditions, for which an accurate deconvolution result cannot be achieved. To show this, we establish a mathematical model to calculate non-monochromatic absorbance values based on imaging equipment typically used in histology and use this simulated data as the ground truth to evaluate the accuracy of colour deconvolution. We show the non-linear characteristics of the absorbance formation and demonstrate how it leads to significant deconvolution errors. In particular, our calculations reveal that polychromatic illumination causes 10-times higher deconvolution errors than sequential monochromatic LED illumination. In conclusion, our model can be used for a quantitative assessment of system components - and also to assess and compare colour deconvolution methods.
Linear and blind colour deconvolution methods are a frequent part of analysis approaches that are based on quantitative and often automated image analysis procedures. For example, CD has been used to detect and classify subcellular protein patterns9. More importantly, however, CD based quantification of IHC tissue microarrays has shown good correlation to manual scoring10,11 and was applied in fully automated manner to score cell death processes12. An open source CD software plugin helped to lower the application barrier so that the method was used in several works to quantify IHC samples13,14,15 and the fact that CD based images were found to achieve the best object segmentation results16 may explain its quick adoption and distribution. In addition, CD is of particular importance for the normalisation of colour in microscopic images, as reviewed elsewhere17. It is used in combination with a colour vector estimation18 and as a part of a machine learning approaches19 to normalise histological images.
We use our model (i) to calculate theoretical RGB colour values for varying concentrations of pure stains, (ii) to study the concentration dependency of stain vectors of pure stains, (iii) to calculate the deconvolution error for double and triple staining, (iv) to visualize the non-linear formation of the absorbance values and (v) to demonstrate the effects of the deconvolution error onto cell detection and feature measurement.
In essence, our study intends to characterize the decomposition error, introduced by linear deconvolution of non-linear data. Quantification of this error is important as inaccuracies created on the CD pre-processing level are likely to propagate into subsequent image analysis procedures, such as object segmentation. Our simulation demonstrates the discrepancy between non-linear signal formation and linear assumption of the CD method. Importantly, our theoretical approach does not comprise error influences from sample preparation and imaging. Therefore, our results describe effects exclusively caused by the methodological error of linear deconvolution of non-linear absorbance signals.
Stain vectors define the target coordinate system for the linear transformation from absorbance into concentration space. Necessarily, they must be specified prior to the deconvolution and can be estimated from samples ideally stained with pure dyes.
These results show that stain vectors, determined from RGB colour images of purely stained biological samples, are influenced by the image region selected for the vector measurement. Hence, stain vectors cannot be determined distinctly from RGB colour images, in particular not under wideband illumination. In consequence, results of linear deconvolution and the deconvolution error strongly depend on the reference conditions.
Shown are concentration results c*out for DAB (a) and HTX (b) and relative deconvolution errors for DAB (c,e) and HTX (d,f). The modelling was based on normalised stain vectors and was calculated for different input concentrations c*in. The deconvolution was performed in the B/G plane projection.
Shown are concentration results c*out for DAB (a) and HTX (b) and relative deconvolution errors for DAB (c,e) and HTX (d,f). The modelling was based on normalised stain vectors and was calculated for different input concentrations c*in. The deconvolution was performed in the B/G plane projection.
The average error values presented above should, however, not be misinterpreted as a general mean deconvolution error. Since the error distribution is extremely non-linear the average values only reflects the error dimension for specific spectral conditions. They rather indicate that the error dimensions are acceptable for a RGB LED illumination, depending on the application requirements, while the error dimensions for the D65 illumination are critical for quantitative applications.
Importantly, as our theoretical model assumes ideal experimental conditions, it only uncovers the methodical weakness of linear deconvolution of non-linear signals, while all sources of experimental errors, e.g. from imaging and sample preparation, that would further increase the total errors, are not included.
To illustrate the deconvolution error under realistic imaging conditions we used DAB and HTX input concentrations to model RGB colour images (example shown in Fig. 5). The input concentrations cinDAB and cinHTX are provided as floating point images. The RGB values were calculated pixel-wise for the spectral conditions of D65/ICX285AQ and RGB LED/ICX285AL using the equations (10) to (12),.
Subsequently we deconvolved them with non-normalised versions of stain vectors from Fig. 2. The deconvolution was calculated by solving the over-determined system of linear equations with an orthogonalization approach based on the well-known QR decomposition. Lastly, the resulting concentration images coutDAB and coutHTX were compared to the input images by calculating the pixel-wise image difference.
These results confirm all of our prior findings, i.e. (i) the general deconvolution error, (ii) the large error for wideband illumination and (iii) the error acceptable for sequential narrow band illumination.
Remarkably, both analyses (Figs 6 and 7) demonstrate that the deconvolution error does not alter image contrast to an extent that obscures morphological cell information. All output images appear visually meaningful and therefore the deconvolution error is often overlooked. This is caused by our visual system and should not mislead us underestimate the error. In consequence, however, this means that a visual inspection of CD images is most likely not affected by the deconvolution error, while a computational qualification most probably is.
based on different ImageJ automatic threshold modes for input concentration images (middle column) and output concentration images derived by QR deconvolution of RGB images which are modelled with D65 illumination/ICX287AQ (left column) and RGB LED illumination/Sony ICX285AL (right column).
In addition, we quantified the number of cells, cell area and stain concentration per cell and per pixel in the thresholded concentration images. In Fig. 9 results of ImageJ cell measurements are shown for the D65 and RGB LED models. The difference between input and output measurements are considerably higher for the D65 compared to the RGB LED data. Overall, for the D65 model more objects have been detected and the mean object area is increased. The relative errors of the mean cell concentrations and mean pixel concentrations give an idea of the error dimensions. The relative concentration error of individual cells is significant higher than the average error values. For the ten largest cells we found DAB concentrations errors of up to 75% for D65 compared to 7% for RGB LED, with a mean of 28% for D65 respectively 1.9% for RGB LED. Again, all prior findings (i.e. (i) the general deconvolution error, (ii) the large error for wideband illumination and (iii) the error acceptable for sequential narrow band illumination) are confirmed by this examination.
35fe9a5643
color deconvolution imagej download
Download https://1niosetincto.blogspot.com/?pu=2x5Ybs
3. Run the Colour Deconvolution2 plugin and select the From ROI option to define example stained areas. Click OK on the plugin dialogue.
The plugin will then ask for rectangular selections of single dye regions. Select ROIs on areas which are all intensely stained (saturated) with only one dye, without empty background regions. Repeat this for other dyes.
If the staining method consists of only two dyes, when prompted for the 3rd selection just right-click. That vector will then be set to 0,0,0 and computed according to the option set for Cross product for Colour 3 as discussed above.
The deconvolution of the test image will take place and the stain separation can be evaluated. The Log window will show something like this (with different values, of course):
Invalid citation arguments: "last"=>"Ruifrok", "first"=>"A.C.", "last2"=>"Johnston", "first2"=>"D.A.", "title"=>"Quantification of histochemical staining by color deconvolution", "journal"=>"Anal. Quant. Cytol. Histol.", "volume"=>"23", "pages"=>"291-299", "year"=>"2001", "PMID"=>"11531144"
Hi,
I uploaded a new version (2.1) of the colour deconvolution2 plugin.
Instructions for ImageJ and Fiji (as well as how to uninstall the one that comes by default with Fiji) can be found here:
-landini-software/colour-deconvolution-2/
Hello! Any tips on using colour deconvolution for analyzing Sirius Red Stains? Or if anyone has a good method for analyzing collagen area from Sirius Red stains in general, that would also be helpful. Thank you!
This version of the Colour Deconvolution plugin works for a single RGB image at the time and only allows a 2-colour deconvolution. Issues in this plugin at this point are that it does not put color...
Colour deconvolution is a method used in diagnostic brightfield microscopy to transform colour images of multiple stained biological samples into images representing the stain concentrations. It is applied by decomposing the absorbance values of stain mixtures into absorbance values of single stains. The method assumes a linear relation between stain concentration and absorbance, which is only valid under monochromatic conditions. Diagnostic applications, in turn, are often performed under polychromatic conditions, for which an accurate deconvolution result cannot be achieved. To show this, we establish a mathematical model to calculate non-monochromatic absorbance values based on imaging equipment typically used in histology and use this simulated data as the ground truth to evaluate the accuracy of colour deconvolution. We show the non-linear characteristics of the absorbance formation and demonstrate how it leads to significant deconvolution errors. In particular, our calculations reveal that polychromatic illumination causes 10-times higher deconvolution errors than sequential monochromatic LED illumination. In conclusion, our model can be used for a quantitative assessment of system components - and also to assess and compare colour deconvolution methods.
Linear and blind colour deconvolution methods are a frequent part of analysis approaches that are based on quantitative and often automated image analysis procedures. For example, CD has been used to detect and classify subcellular protein patterns9. More importantly, however, CD based quantification of IHC tissue microarrays has shown good correlation to manual scoring10,11 and was applied in fully automated manner to score cell death processes12. An open source CD software plugin helped to lower the application barrier so that the method was used in several works to quantify IHC samples13,14,15 and the fact that CD based images were found to achieve the best object segmentation results16 may explain its quick adoption and distribution. In addition, CD is of particular importance for the normalisation of colour in microscopic images, as reviewed elsewhere17. It is used in combination with a colour vector estimation18 and as a part of a machine learning approaches19 to normalise histological images.
We use our model (i) to calculate theoretical RGB colour values for varying concentrations of pure stains, (ii) to study the concentration dependency of stain vectors of pure stains, (iii) to calculate the deconvolution error for double and triple staining, (iv) to visualize the non-linear formation of the absorbance values and (v) to demonstrate the effects of the deconvolution error onto cell detection and feature measurement.
In essence, our study intends to characterize the decomposition error, introduced by linear deconvolution of non-linear data. Quantification of this error is important as inaccuracies created on the CD pre-processing level are likely to propagate into subsequent image analysis procedures, such as object segmentation. Our simulation demonstrates the discrepancy between non-linear signal formation and linear assumption of the CD method. Importantly, our theoretical approach does not comprise error influences from sample preparation and imaging. Therefore, our results describe effects exclusively caused by the methodological error of linear deconvolution of non-linear absorbance signals.
Stain vectors define the target coordinate system for the linear transformation from absorbance into concentration space. Necessarily, they must be specified prior to the deconvolution and can be estimated from samples ideally stained with pure dyes.
These results show that stain vectors, determined from RGB colour images of purely stained biological samples, are influenced by the image region selected for the vector measurement. Hence, stain vectors cannot be determined distinctly from RGB colour images, in particular not under wideband illumination. In consequence, results of linear deconvolution and the deconvolution error strongly depend on the reference conditions.
Shown are concentration results c*out for DAB (a) and HTX (b) and relative deconvolution errors for DAB (c,e) and HTX (d,f). The modelling was based on normalised stain vectors and was calculated for different input concentrations c*in. The deconvolution was performed in the B/G plane projection.
Shown are concentration results c*out for DAB (a) and HTX (b) and relative deconvolution errors for DAB (c,e) and HTX (d,f). The modelling was based on normalised stain vectors and was calculated for different input concentrations c*in. The deconvolution was performed in the B/G plane projection.
The average error values presented above should, however, not be misinterpreted as a general mean deconvolution error. Since the error distribution is extremely non-linear the average values only reflects the error dimension for specific spectral conditions. They rather indicate that the error dimensions are acceptable for a RGB LED illumination, depending on the application requirements, while the error dimensions for the D65 illumination are critical for quantitative applications.
Importantly, as our theoretical model assumes ideal experimental conditions, it only uncovers the methodical weakness of linear deconvolution of non-linear signals, while all sources of experimental errors, e.g. from imaging and sample preparation, that would further increase the total errors, are not included.
To illustrate the deconvolution error under realistic imaging conditions we used DAB and HTX input concentrations to model RGB colour images (example shown in Fig. 5). The input concentrations cinDAB and cinHTX are provided as floating point images. The RGB values were calculated pixel-wise for the spectral conditions of D65/ICX285AQ and RGB LED/ICX285AL using the equations (10) to (12),.
Subsequently we deconvolved them with non-normalised versions of stain vectors from Fig. 2. The deconvolution was calculated by solving the over-determined system of linear equations with an orthogonalization approach based on the well-known QR decomposition. Lastly, the resulting concentration images coutDAB and coutHTX were compared to the input images by calculating the pixel-wise image difference.
These results confirm all of our prior findings, i.e. (i) the general deconvolution error, (ii) the large error for wideband illumination and (iii) the error acceptable for sequential narrow band illumination.
Remarkably, both analyses (Figs 6 and 7) demonstrate that the deconvolution error does not alter image contrast to an extent that obscures morphological cell information. All output images appear visually meaningful and therefore the deconvolution error is often overlooked. This is caused by our visual system and should not mislead us underestimate the error. In consequence, however, this means that a visual inspection of CD images is most likely not affected by the deconvolution error, while a computational qualification most probably is.
based on different ImageJ automatic threshold modes for input concentration images (middle column) and output concentration images derived by QR deconvolution of RGB images which are modelled with D65 illumination/ICX287AQ (left column) and RGB LED illumination/Sony ICX285AL (right column).
In addition, we quantified the number of cells, cell area and stain concentration per cell and per pixel in the thresholded concentration images. In Fig. 9 results of ImageJ cell measurements are shown for the D65 and RGB LED models. The difference between input and output measurements are considerably higher for the D65 compared to the RGB LED data. Overall, for the D65 model more objects have been detected and the mean object area is increased. The relative errors of the mean cell concentrations and mean pixel concentrations give an idea of the error dimensions. The relative concentration error of individual cells is significant higher than the average error values. For the ten largest cells we found DAB concentrations errors of up to 75% for D65 compared to 7% for RGB LED, with a mean of 28% for D65 respectively 1.9% for RGB LED. Again, all prior findings (i.e. (i) the general deconvolution error, (ii) the large error for wideband illumination and (iii) the error acceptable for sequential narrow band illumination) are confirmed by this examination.
35fe9a5643
0 new messages