Hi,
I have tissue micro-array scanned images which I would like to analyze with IHCprofiler (imageJ plugin). I have close to 1000 arrays to process so automated analysis seems the way to go.
All of the images I have tried, including those which should be positive, are returning as low positive or negative with the Imagej IHC profiler plugin.
I would appreciate if anyone can offer some advice or assistance.
My current protocol is:
>open scanned TMA slide in NDP viewer, zoom to core of interest and export image as jpeg.
>open jpeg in imagej, crop to a region of interest to exclude background.
> Plugins / IHC Profiler / IHC profiler / cytoplasmic stain / H DAB
Any advice on this is very much appreciated.
Many thanks,
Hi Alison,
Another student in our group was working with this program recently. My understanding of "positive/low positive" categories is that they are arbitrary divisions of the intensity values - so they may not be appropriate objective metrics depending on your analytical pipeline. Think of them as "course histogram bins." I suspect your analysis lacks sensitivity because these bins are too coarse.
What the plugin does is print a distribution of intensity values and then report the relative presence of values across four bins/categories (none, low, positive, and high positive). So effectively, those categories give you a histogram of the data in four bins (for convenience and comparison to semi-quantitative "scoring" methods commonly used in histology). There's no reason to limit analysis to those categories; one could easily bin the data into smaller categories and gain greater resolution. Even better would be to compare the pixel distributions using nonparametric statistical tests to see if they come from the same underlying distribution or not. The data can be accessed with the "list" functionality in the ImageJ plot window.
The accompanying paper to the plugin also recommends using a threshold on your images in its example workflow. If you aren't using a threshold to minimize the influence of background pixel values, I'd suggest doing that first and seeing if it removes noise from your histograms. In principle, you could use your negative controls to "zero" out the background if you have clear morphological features, such that only those features are being measured and categorized.
A final note is that the results are sensitive to magnification, so the distributions won't hold across different zooms unless the pixel intensities are homogeneously distributed at all window sizes (unlikely when biological structures are involved). This is probably important to consider if you sample large fields with several ROIs.
Good luck!
Article IHC Profiler: An Open Source Plugin for the Quantitative Eva...
Hi Alison,
I have found the same problem. I'm analysing soft tissue slide and some of positive control slide not result positive with the IHC Profiler. How did you solve it?
Hi Alison,
I also experienced the same thing in the past. Do you do an individual analysis of each image? If so, sometimes individual analyzes can produce errors like this. I also don't understand why this error can occur. However, I have a solution to avoid this error with batch processing.
Batch processing is a mechanism within ImageJ that allows users to perform multiple image analysis at once. To do so, try to follow my protocol:
1. You need to prepare the image that you will process. This means that you need to "crop to a region of interest to exclude background" on all images first.
2. After all the images are ready. Put the image in one or more folders according to your needs.
3. Click Process> batch> macro and the batch process window will appear.
4. Select the input folder containing the image you wish to analyze. Remember, you can only batch process one folder at a time! So if you are working in multiple folders, you need to do batch processing equal to the number of your folders.
5. Select the output folder where you want it. Basically, this output folder will contain the logs and histograms of the analysis of all the images in your input.
6. Set the output format according to your needs. I usually don't change the settings for "output format", "add macro code", and "file name contain".
7. Under the file name contains, there will be a large empty space. It's where you put your batch processing code. Each protocol has a different code. I have coded your protocol as the following:
run ("IHC Profiler", "mode = [Cytoplasmic Stained Image] vectors = [H DAB]");
8. After that click process, than the log and histogram window will start batch processing all the images in your input file.
This batch processing has two advantages. The first is to streamline your work on image analysis. Second, for some reason, if you use this batch processing, the error that you asked about earlier will not occur.
I hope my answer helps. I've also included my publications if you'd like to read more ( https://www.phcogj.com/article/1326 ). Contact me if you have further questions. Cheers!
Elvan Wiyarta Thank you so much for providing that solution! I had the same problem with the individual processing, but your advice worked perfectly.
I do however have another question... After you got those values (in your own article now) you converted it to H-score with the provided math equation. Why did you have to do that?
I have my intensity values now, but I am unsure what to do with it next? For my study, I stained my slides with the antibody for VEFG to determine angiogenesis, and I have 2 different treatment groups. I just want the staining intensity to confirm upregulation of VEGF (thus also angiogenesis aka increased formation of blood vessels).
Your advice would be much appreciated!
- Badges
- Science method
- Similar topics
- Biological Science
- Histology
- Staining